Deep learning

Source: Wikipedia, the free encyclopedia.
For the TV series episode, see Deep Learning (South Park).
Representing images on multiple layers of abstraction in deep learning
Representing images on multiple layers of abstraction in deep learning[1]
Part of a series on
Artificial intelligence
Major goals
Approaches
Applications
Philosophy
History
Glossary

Deep learning is the subset of

semi-supervised or unsupervised.[2]

Deep-learning architectures such as

medical image analysis, climate science, material inspection and board game programs, where they have produced results comparable to and in some cases surpassing human expert performance.[3][4][5]

Artificial neural networks were inspired by information processing and distributed communication nodes in biological systems. ANNs have various differences from biological brains. Specifically, artificial neural networks tend to be static and symbolic, while the biological brain of most living organisms is dynamic (plastic) and analog.[6][7] ANNs are generally seen as low quality models for brain function.[8]

Definition

Deep learning is a class of

image processing
, lower layers may identify edges, while higher layers may identify the concepts relevant to a human, such as digits, letters, or faces.

From another angle to view deep learning, deep learning refers to "computer-simulate" or "automate" human learning processes from a source (e.g., an image of dogs) to a learned object (dogs). Therefore, a notion coined as "deeper" learning or "deepest" learning[10] makes sense. The deepest learning refers to the fully automatic learning from a source to a final learned object. A deeper learning thus refers to a mixed learning process: a human learning process from a source to a learned semi-object, followed by a computer learning process from the human learned semi-object to a final learned object.

Overview

Most modern deep learning models are based on multi-layered

transformers, although they can also include propositional formulas or latent variables organized layer-wise in deep generative models such as the nodes in deep belief networks and deep Boltzmann machines.[11]

In deep learning, each level learns to transform its input data into a slightly more abstract and composite representation. In an image recognition application, the raw input may be a matrix of pixels; the first representational layer may abstract the pixels and encode edges; the second layer may compose and encode arrangements of edges; the third layer may encode a nose and eyes; and the fourth layer may recognize that the image contains a face. Importantly, a deep learning process can learn which features to optimally place in which level on its own. This does not eliminate the need for hand-tuning; for example, varying numbers of layers and layer sizes can provide different degrees of abstraction.[12][13]

The word "deep" in "deep learning" refers to the number of layers through which the data is transformed. More precisely, deep learning systems have a substantial credit assignment path (CAP) depth. The CAP is the chain of transformations from input to output. CAPs describe potentially causal connections between input and output. For a feedforward neural network, the depth of the CAPs is that of the network and is the number of hidden layers plus one (as the output layer is also parameterized). For recurrent neural networks, in which a signal may propagate through a layer more than once, the CAP depth is potentially unlimited.[14] No universally agreed-upon threshold of depth divides shallow learning from deep learning, but most researchers agree that deep learning involves CAP depth higher than 2. CAP of depth 2 has been shown to be a universal approximator in the sense that it can emulate any function.[15] Beyond that, more layers do not add to the function approximator ability of the network. Deep models (CAP > 2) are able to extract better features than shallow models and hence, extra layers help in learning the features effectively.

Deep learning architectures can be constructed with a greedy layer-by-layer method.[16] Deep learning helps to disentangle these abstractions and pick out which features improve performance.[12]

For

principal components
, and derive layered structures that remove redundancy in representation.

Deep learning algorithms can be applied to unsupervised learning tasks. This is an important benefit because unlabeled data are more abundant than the labeled data. Examples of deep structures that can be trained in an unsupervised manner are deep belief networks.[12][17]

Machine learning models are now adept at identifying complex patterns in financial market data. Due to the benefits of artificial intelligence, investors are increasingly utilizing deep learning techniques to forecast and analyze trends in stock and foreign exchange markets.[18]

Interpretations

Deep neural networks are generally interpreted in terms of the universal approximation theorem[19][20][21][22][23] or probabilistic inference.[24][9][12][14][25]

The classic universal approximation theorem concerns the capacity of

rectified linear unit.[26][27]

The universal approximation theorem for

ReLU activation is strictly larger than the input dimension, then the network can approximate any Lebesgue integrable function
; if the width is smaller or equal to the input dimension, then a deep neural network is not a universal approximator.

The

dropout as regularizer in neural networks. The probabilistic interpretation was introduced by researchers including Hopfield, Widrow and Narendra and popularized in surveys such as the one by Bishop.[28]

History

There are two types of artificial neural network (ANN):

language processing
.

Charles Tappert writes that Frank Rosenblatt developed and explored all of the basic ingredients of the deep learning systems of today,[33] referring to Rosenblatt's 1962 book[34] which introduced multilayer perceptron (MLP) with 3 layers: an input layer, a hidden layer with randomized weights that did not learn, and an output layer. It also introduced variants, including a version with four-layer perceptrons where the last two layers have learned weights (and thus a proper multilayer perceptron).[34]: section 16  In addition, term deep learning was proposed in 1986 by Rina Dechter[35] although the history of its appearance is apparently more complicated.[36]

The first general, working learning algorithm for supervised, deep, feedforward, multilayer perceptrons was published by Alexey Ivakhnenko and Lapa in 1967.[37] A 1971 paper described a deep network with eight layers trained by the group method of data handling.[38]

The first deep learning

internal representations to classify non-linearily separable pattern classes.[31] In 1987 Matthew Brand reported that wide 12-layer nonlinear perceptrons could be fully end-to-end trained to reproduce logic functions of nontrivial circuit depth via gradient descent on small batches of random input/output samples, but concluded that training time on contemporary hardware (sub-megaflop computers) made the technique impractical, and proposed using fixed random early layers as an input hash for a single modifiable layer.[41] Instead, subsequent developments in hardware and hyperparameter tunings have made end-to-end stochastic gradient descent
the currently dominant training technique.

In 1970, Seppo Linnainmaa published the reverse mode of automatic differentiation of discrete connected networks of nested differentiable functions.[42][43][44] This became known as backpropagation.[14] It is an efficient application of the chain rule derived by Gottfried Wilhelm Leibniz in 1673[45] to networks of differentiable nodes.[31] The terminology "back-propagating errors" was actually introduced in 1962 by Rosenblatt,

David E. Rumelhart et al. published an experimental analysis of the technique.[49]

Deep learning architectures for convolutional neural networks (CNNs) with convolutional layers and downsampling layers began with the Neocognitron introduced by Kunihiko Fukushima in 1980.[50] In 1969, he also introduced the ReLU (rectified linear unit) activation function.[26][31] The rectifier has become the most popular activation function for CNNs and deep learning in general.[51] CNNs have become an essential tool for computer vision.

The term Deep Learning was introduced to the machine learning community by Rina Dechter in 1986,[35] and to artificial neural networks by Igor Aizenberg and colleagues in 2000, in the context of Boolean threshold neurons.[52][53]

In 1988, Wei Zhang et al. applied the backpropagation algorithm to a convolutional neural network (a simplified Neocognitron with convolutional interconnections between the image feature layers and the last fully connected layer) for alphabet recognition. They also proposed an implementation of the CNN with an optical computing system.[54][55] In 1989, Yann LeCun et al. applied backpropagation to a CNN with the purpose of recognizing handwritten ZIP codes on mail. While the algorithm worked, training required 3 days.[56] Subsequently, Wei Zhang, et al. modified their model by removing the last fully connected layer and applied it for medical image object segmentation in 1991[57] and breast cancer detection in mammograms in 1994.[58] LeNet-5 (1998), a 7-level CNN by Yann LeCun et al.,[59] that classifies digits, was applied by several banks to recognize hand-written numbers on checks digitized in 32x32 pixel images.

In the 1980s, backpropagation did not work well for deep learning with long credit assignment paths. To overcome this problem,

internal representations at multiple self-organizing time scales. This can substantially facilitate downstream deep learning. The RNN hierarchy can be collapsed into a single RNN, by distilling a higher level chunker network into a lower level automatizer network.[60][31] In 1993, a chunker solved a deep learning task whose depth exceeded 1000.[61]

In 1992,

Transformer or a Transformer with linearized self-attention[63][64][31] (save for a normalization operator). It learns internal spotlights of attention:[65] a slow feedforward neural network learns by gradient descent to control the fast weights of another neural network through outer products of self-generated activation patterns FROM and TO (which are now called key and value for self-attention).[63]
This fast weight attention mapping is applied to a query pattern.

The modern

Transformer was introduced by Ashish Vaswani et al. in their 2017 paper "Attention Is All You Need".[66]
It combines this with a
softmax operator and a projection matrix.[31]
Transformers have increasingly become the model of choice for natural language processing.[67] Many modern large language models such as ChatGPT, GPT-4, and BERT use it. Transformers are also increasingly being used in computer vision.[68]

In 1991, Jürgen Schmidhuber also published adversarial neural networks that contest with each other in the form of a zero-sum game, where one network's gain is the other network's loss.[69][70][71] The first network is a generative model that models a probability distribution over output patterns. The second network learns by gradient descent to predict the reactions of the environment to these patterns. This was called "artificial curiosity". In 2014, this principle was used in a generative adversarial network (GAN) by Ian Goodfellow et al.[72] Here the environmental reaction is 1 or 0 depending on whether the first network's output is in a given set. This can be used to create realistic deepfakes.[73] Excellent image quality is achieved by Nvidia's StyleGAN (2018)[74] based on the Progressive GAN by Tero Karras et al.[75] Here the GAN generator is grown from small to large scale in a pyramidal fashion.

LSTM has become the most cited neural network of the 20th century.[31]
In 2015, Rupesh Kumar Srivastava, Klaus Greff, and Schmidhuber used
LSTM principles to create the Highway network, a feedforward neural network with hundreds of layers, much deeper than previous networks.[80][81] 7 months later, Kaiming He, Xiangyu Zhang; Shaoqing Ren, and Jian Sun won the ImageNet 2015 competition with an open-gated or gateless Highway network variant called Residual neural network.[82] This has become the most cited neural network of the 21st century.[31]

In 1994, André de Carvalho, together with Mike Fairhurst and David Bisset, published experimental results of a multi-layer boolean neural network, also known as a weightless neural network, composed of a 3-layers self-organising feature extraction neural network module (SOFT) followed by a multi-layer classification neural network module (GSN), which were independently trained. Each layer in the feature extraction module extracted features with growing complexity regarding the previous layer.[83]

In 1995, Brendan Frey demonstrated that it was possible to train (over two days) a network containing six fully connected layers and several hundred hidden units using the wake-sleep algorithm, co-developed with Peter Dayan and Hinton.[84]

Since 1997, Sven Behnke extended the feed-forward hierarchical convolutional approach in the Neural Abstraction Pyramid[85] by lateral and backward connections in order to flexibly incorporate context into decisions and iteratively resolve local ambiguities.

Simpler models that use task-specific handcrafted features such as Gabor filters and support vector machines (SVMs) were a popular choice in the 1990s and 2000s, because of artificial neural networks' computational cost and a lack of understanding of how the brain wires its biological networks.

Both shallow and deep learning (e.g., recurrent nets) of ANNs for speech recognition have been explored for many years.[86][87][88] These methods never outperformed non-uniform internal-handcrafting Gaussian mixture model/Hidden Markov model (GMM-HMM) technology based on generative models of speech trained discriminatively.[89] Key difficulties have been analyzed, including gradient diminishing[76] and weak temporal correlation structure in neural predictive models.[90][91] Additional difficulties were the lack of training data and limited computing power. Most speech recognition researchers moved away from neural nets to pursue generative modeling. An exception was at SRI International in the late 1990s. Funded by the US government's NSA and DARPA, SRI studied deep neural networks (DNNs) in speech and speaker recognition. The speaker recognition team led by Larry Heck reported significant success with deep neural networks in speech processing in the 1998 National Institute of Standards and Technology Speaker Recognition evaluation.[92] The SRI deep neural network was then deployed in the Nuance Verifier, representing the first major industrial application of deep learning.[93] The principle of elevating "raw" features over hand-crafted optimization was first explored successfully in the architecture of deep autoencoder on the "raw" spectrogram or linear filter-bank features in the late 1990s,[93] showing its superiority over the Mel-Cepstral features that contain stages of fixed transformation from spectrograms. The raw features of speech, waveforms, later produced excellent larger-scale results.[94]

Speech recognition was taken over by

LSTM. In 2003, LSTM started to become competitive with traditional speech recognizers on certain tasks.[95] In 2006, Alex Graves, Santiago Fernández, Faustino Gomez, and Schmidhuber combined it with connectionist temporal classification (CTC)[96] in stacks of LSTM RNNs.[97] In 2015, Google's speech recognition reportedly experienced a dramatic performance jump of 49% through CTC-trained LSTM, which they made available through Google Voice Search.[98]

The impact of deep learning in industry began in the early 2000s, when CNNs already processed an estimated 10% to 20% of all the checks written in the US, according to Yann LeCun.[99] Industrial applications of deep learning to large-scale speech recognition started around 2010.

In 2006, publications by Geoff Hinton, Ruslan Salakhutdinov, Osindero and Teh[100][101][102] showed how a many-layered feedforward neural network could be effectively pre-trained one layer at a time, treating each layer in turn as an unsupervised restricted Boltzmann machine, then fine-tuning it using supervised backpropagation.[103] The papers referred to learning for deep belief nets.

The 2009 NIPS Workshop on Deep Learning for Speech Recognition was motivated by the limitations of deep generative models of speech, and the possibility that given more capable hardware and large-scale data sets that deep neural nets might become practical. It was believed that pre-training DNNs using generative models of deep belief nets (DBN) would overcome the main difficulties of neural nets. However, it was discovered that replacing pre-training with large amounts of training data for straightforward backpropagation when using DNNs with large, context-dependent output layers produced error rates dramatically lower than then-state-of-the-art Gaussian mixture model (GMM)/Hidden Markov Model (HMM) and also than more-advanced generative model-based systems.[104] The nature of the recognition errors produced by the two types of systems was characteristically different,[105] offering technical insights into how to integrate deep learning into the existing highly efficient, run-time speech decoding system deployed by all major speech recognition systems.[9][106][107] Analysis around 2009–2010, contrasting the GMM (and other generative speech models) vs. DNN models, stimulated early industrial investment in deep learning for speech recognition.[105] That analysis was done with comparable performance (less than 1.5% in error rate) between discriminative DNNs and generative models.[104][105][108] In 2010, researchers extended deep learning from TIMIT to large vocabulary speech recognition, by adopting large output layers of the DNN based on context-dependent HMM states constructed by decision trees.[109][110][111][106]

Deep learning is part of state-of-the-art systems in various disciplines, particularly computer vision and

LSTM.[78][98][113][114][115]
but are more successful in computer vision.

Advances in hardware have driven renewed interest in deep learning. In 2009, Nvidia was involved in what was called the "big bang" of deep learning, "as deep-learning neural networks were trained with Nvidia graphics processing units (GPUs)".[116] That year, Andrew Ng determined that GPUs could increase the speed of deep-learning systems by about 100 times.[117] In particular, GPUs are well-suited for the matrix/vector computations involved in machine learning.[118][119][120] GPUs speed up training algorithms by orders of magnitude, reducing running times from weeks to days.[121][122] Further, specialized hardware and algorithm optimizations can be used for efficient processing of deep learning models.[123]

Deep learning revolution

How deep learning is a subset of machine learning and how machine learning is a subset of artificial intelligence (AI)

In the late 2000s, deep learning started to outperform other methods in machine learning competitions. In 2009, a long short-term memory trained by connectionist temporal classification (Alex Graves, Santiago Fernández, Faustino Gomez, and Jürgen Schmidhuber, 2006)[96] was the first RNN to win pattern recognition contests, winning three competitions in connected handwriting recognition.[124][14] Google later used CTC-trained LSTM for speech recognition on the smartphone.[125][98]

Significant impacts in image or object recognition were felt from 2011 to 2012. Although CNNs trained by backpropagation had been around for decades,

ImageNet competition
by a significant margin over shallow machine learning methods. The VGG-16 network by Karen Simonyan and Andrew Zisserman[129] further reduced the error rate and won the ImageNet 2014 competition, following a similar trend in large-scale speech recognition.

Image classification was then extended to the more challenging task of generating descriptions (captions) for images, often as a combination of CNNs and LSTMs.[130][131][132]

In 2012, a team led by George E. Dahl won the "Merck Molecular Activity Challenge" using multi-task deep neural networks to predict the

FDA and NCATS.[135][136][137]

In 2016, Roger Parloff mentioned a "deep learning revolution" that has transformed the AI industry.[138]

In March 2019, Yoshua Bengio, Geoffrey Hinton and Yann LeCun were awarded the Turing Award for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing.

Neural networks

Main article:
Artificial neural network
false positive
result for sea urchin.
In reality, textures and outlines would not be represented by single nodes, but rather by associated weight patterns of multiple nodes.

Artificial neural networks (ANNs) or

rule-based programming
.

An ANN is based on a collection of connected units called

real numbers
, typically between 0 and 1. Neurons and synapses may also have a weight that varies as learning proceeds, which can increase or decrease the strength of the signal that it sends downstream.

Typically, neurons are organized in layers. Different layers may perform different kinds of transformations on their inputs. Signals travel from the first (input), to the last (output) layer, possibly after traversing the layers multiple times.

The original goal of the neural network approach was to solve problems in the same way that a human brain would. Over time, attention focused on matching specific mental abilities, leading to deviations from biology such as backpropagation, or passing information in the reverse direction and adjusting the network to reflect that information.

Neural networks have been used on a variety of tasks, including computer vision, speech recognition, machine translation, social network filtering, playing board and video games and medical diagnosis.

As of 2017, neural networks typically have a few thousand to a few million units and millions of connections. Despite this number being several order of magnitude less than the number of neurons on a human brain, these networks can perform many tasks at a level beyond that of humans (e.g., recognizing faces, or playing "Go"[140]).

Deep neural networks

A deep neural network (DNN) is an artificial neural network with multiple layers between the input and output layers.[11][14] There are different types of neural networks but they always consist of the same components: neurons, synapses, weights, biases, and functions.[141] These components as a whole function in a way that mimics functions of the human brain, and can be trained like any other ML algorithm.[citation needed]

For example, a DNN that is trained to recognize dog breeds will go over the given image and calculate the probability that the dog in the image is a certain breed. The user can review the results and select which probabilities the network should display (above a certain threshold, etc.) and return the proposed label. Each mathematical manipulation as such is considered a layer,[citation needed] and complex DNN have many layers, hence the name "deep" networks.

DNNs can model complex non-linear relationships. DNN architectures generate compositional models where the object is expressed as a layered composition of

multivariate polynomials are exponentially easier to approximate with DNNs than with shallow networks.[143]

Deep architectures include many variants of a few basic approaches. Each architecture has found success in specific domains. It is not always possible to compare the performance of multiple architectures, unless they have been evaluated on the same data sets.

DNNs are typically feedforward networks in which data flows from the input layer to the output layer without looping back. At first, the DNN creates a map of virtual neurons and assigns random numerical values, or "weights", to connections between them. The weights and inputs are multiplied and return an output between 0 and 1. If the network did not accurately recognize a particular pattern, an algorithm would adjust the weights.[144] That way the algorithm can make certain parameters more influential, until it determines the correct mathematical manipulation to fully process the data.

Recurrent neural networks, in which data can flow in any direction, are used for applications such as language modeling.[145][146][147][148][149] Long short-term memory is particularly effective for this use.[78][150]

Convolutional neural networks (CNNs) are used in computer vision.[151] CNNs also have been applied to acoustic modeling for automatic speech recognition (ASR).[152]

Challenges

As with ANNs, many issues can arise with naively trained DNNs. Two common issues are overfitting and computation time.

DNNs are prone to overfitting because of the added layers of abstraction, which allow them to model rare dependencies in the training data.

weight decay
(-regularization) or sparsity (-regularization) can be applied during training to combat overfitting.
dropout regularization randomly omits units from the hidden layers during training. This helps to exclude rare dependencies.[154] Finally, data can be augmented via methods such as cropping and rotating such that smaller training sets can be increased in size to reduce the chances of overfitting.[155]

DNNs must consider many training parameters, such as the size (number of layers and number of units per layer), the

batching (computing the gradient on several training examples at once rather than individual examples)[156] speed up computation. Large processing capabilities of many-core architectures (such as GPUs or the Intel Xeon Phi) have produced significant speedups in training, because of the suitability of such processing architectures for the matrix and vector computations.[157][158]

Alternatively, engineers may look for other types of neural networks with more straightforward and convergent training algorithms. CMAC (cerebellar model articulation controller) is one such kind of neural network. It doesn't require learning rates or randomized initial weights. The training process can be guaranteed to converge in one step with a new batch of data, and the computational complexity of the training algorithm is linear with respect to the number of neurons involved.[159][160]

Hardware

Since the 2010s, advances in both machine learning algorithms and

GPUs), often with AI-specific enhancements, had displaced CPUs as the dominant method of training large-scale commercial cloud AI.[162] OpenAI estimated the hardware computation used in the largest deep learning projects from AlexNet (2012) to AlphaZero (2017), and found a 300,000-fold increase in the amount of computation required, with a doubling-time trendline of 3.4 months.[163][164]

Special

tensor processing units (TPU) in the Google Cloud Platform.[166] Cerebras Systems has also built a dedicated system to handle large deep learning models, the CS-2, based on the largest processor in the industry, the second-generation Wafer Scale Engine (WSE-2).[167][168]

Atomically thin

semiconductors
are considered promising for energy-efficient deep learning hardware where the same basic device structure is used for both logic operations and data storage. In 2020, Marega et al. published experiments with a large-area active channel material for developing logic-in-memory devices and circuits based on
floating-gate field-effect transistors (FGFETs).[169]

In 2021, J. Feldmann et al. proposed an integrated

hardware accelerator for parallel convolutional processing.[170] The authors identify two key advantages of integrated photonics over its electronic counterparts: (1) massively parallel data transfer through wavelength division multiplexing in conjunction with frequency combs, and (2) extremely high data modulation speeds.[170] Their system can execute trillions of multiply-accumulate operations per second, indicating the potential of integrated photonics in data-heavy AI applications.[170]

Applications

Automatic speech recognition

Main article: Speech recognition

Large-scale automatic speech recognition is the first and most convincing successful case of deep learning. LSTM RNNs can learn "Very Deep Learning" tasks[14] that involve multi-second intervals containing speech events separated by thousands of discrete time steps, where one time step corresponds to about 10 ms. LSTM with forget gates[150] is competitive with traditional speech recognizers on certain tasks.[95]

The initial success in speech recognition was based on small-scale recognition tasks based on TIMIT. The data set contains 630 speakers from eight major dialects of American English, where each speaker reads 10 sentences.[171] Its small size lets many configurations be tried. More importantly, the TIMIT task concerns phone-sequence recognition, which, unlike word-sequence recognition, allows weak phone bigram language models. This lets the strength of the acoustic modeling aspects of speech recognition be more easily analyzed. The error rates listed below, including these early results and measured as percent phone error rates (PER), have been summarized since 1991.

Method Percent phone
error rate (PER) (%)
Randomly Initialized RNN[172] 26.1
Bayesian Triphone GMM-HMM 25.6
Hidden Trajectory (Generative) Model 24.8
Monophone Randomly Initialized DNN 23.4
Monophone DBN-DNN 22.4
Triphone GMM-HMM with BMMI Training 21.7
Monophone DBN-DNN on fbank 20.7
Convolutional DNN[173] 20.0
Convolutional DNN w. Heterogeneous Pooling 18.7
Ensemble DNN/CNN/RNN[174] 18.3
Bidirectional LSTM 17.8
Hierarchical Convolutional Deep Maxout Network[175] 16.5

The debut of DNNs for speaker recognition in the late 1990s and speech recognition around 2009-2011 and of LSTM around 2003–2007, accelerated progress in eight major areas:[9][108][106]

  • Scale-up/out and accelerated DNN training and decoding
  • Sequence discriminative training
  • Feature processing by deep models with solid understanding of the underlying mechanisms
  • Adaptation of DNNs and related deep models
  • Multi-task and transfer learning by DNNs and related deep models
  • CNNs and how to design them to best exploit domain knowledge of speech
  • RNN and its rich LSTM variants
  • Other types of deep models including tensor-based models and integrated deep generative/discriminative models.

All major commercial speech recognition systems (e.g., Microsoft

Cortana, Xbox, Skype Translator, Amazon Alexa, Google Now, Apple Siri, Baidu and iFlyTek voice search, and a range of Nuance speech products, etc.) are based on deep learning.[9][176][177]

Image recognition

Main article: Computer vision

A common evaluation set for image classification is the MNIST database data set. MNIST is composed of handwritten digits and includes 60,000 training examples and 10,000 test examples. As with TIMIT, its small size lets users test multiple configurations. A comprehensive list of results on this set is available.[178]

Deep learning-based image recognition has become "superhuman", producing more accurate results than human contestants. This first occurred in 2011 in recognition of traffic signs, and in 2014, with recognition of human faces.[179][180]

Deep learning-trained vehicles now interpret 360° camera views.[181] Another example is Facial Dysmorphology Novel Analysis (FDNA) used to analyze cases of human malformation connected to a large database of genetic syndromes.

Visual art processing

Visual art processing of Jimmy Wales in France, with the style of Munch's "The Scream" applied using neural style transfer

Closely related to the progress that has been made in image recognition is the increasing application of deep learning techniques to various visual art tasks. DNNs have proven themselves capable, for example, of

Natural language processing

Main article: Natural language processing

Neural networks have been used for implementing language models since the early 2000s.[145] LSTM helped to improve machine translation and language modeling.[146][147][148]

Other key techniques in this field are negative sampling

probabilistic context free grammar (PCFG) implemented by an RNN.[185] Recursive auto-encoders built atop word embeddings can assess sentence similarity and detect paraphrasing.[185] Deep neural architectures provide the best results for constituency parsing,[186] sentiment analysis,[187] information retrieval,[188][189] spoken language understanding,[190] machine translation,[146][191] contextual entity linking,[191] writing style recognition,[192] named-entity recognition (token classification),[193] text classification, and others.[194]

Recent developments generalize word embedding to sentence embedding.

Google Translate (GT) uses a large end-to-end long short-term memory (LSTM) network.[195][196][197][198] Google Neural Machine Translation (GNMT) uses an example-based machine translation method in which the system "learns from millions of examples".[196] It translates "whole sentences at a time, rather than pieces". Google Translate supports over one hundred languages.[196] The network encodes the "semantics of the sentence rather than simply memorizing phrase-to-phrase translations".[196][199] GT uses English as an intermediate between most language pairs.[199]

Drug discovery and toxicology

For more information, see Drug discovery and Toxicology.

A large percentage of candidate drugs fail to win regulatory approval. These failures are caused by insufficient efficacy (on-target effect), undesired interactions (off-target effects), or unanticipated

off-targets, and toxic effects of environmental chemicals in nutrients, household products and drugs.[135][136][137]

AtomNet is a deep learning system for structure-based

Ebola virus[203] and multiple sclerosis.[204][203]

In 2017 graph neural networks were used for the first time to predict various properties of molecules in a large toxicology data set.[205] In 2019, generative neural networks were used to produce molecules that were validated experimentally all the way into mice.[206][207]

Customer relationship management

Main article: Customer relationship management

RFM variables. The estimated value function was shown to have a natural interpretation as customer lifetime value.[208]

Recommendation systems

Main article: Recommender system

Recommendation systems have used deep learning to extract meaningful features for a latent factor model for content-based music and journal recommendations.[209][210] Multi-view deep learning has been applied for learning user preferences from multiple domains.[211] The model uses a hybrid collaborative and content-based approach and enhances recommendations in multiple tasks.

Bioinformatics

Main article: Bioinformatics

An autoencoder ANN was used in bioinformatics, to predict gene ontology annotations and gene-function relationships.[212]

In medical informatics, deep learning was used to predict sleep quality based on data from wearables[213] and predictions of health complications from electronic health record data.[214]

Deep Neural Network Estimations

Deep neural networks can be used to estimate the entropy of a

Softmax layer with number of nodes that is equal to the alphabet size of Y. NJEE uses continuously differentiable activation functions, such that the conditions for the universal approximation theorem holds. It is shown that this method provides a strongly consistent estimator and outperforms other methods in case of large alphabet sizes.[215]

Medical image analysis

Deep learning has been shown to produce competitive results in medical application such as cancer cell classification, lesion detection, organ segmentation and image enhancement.[216][217] Modern deep learning tools demonstrate the high accuracy of detecting various diseases and the helpfulness of their use by specialists to improve the diagnosis efficiency.[218][219]

Mobile advertising

Finding the appropriate mobile audience for mobile advertising is always challenging, since many data points must be considered and analyzed before a target segment can be created and used in ad serving by any ad server.[220] Deep learning has been used to interpret large, many-dimensioned advertising datasets. Many data points are collected during the request/serve/click internet advertising cycle. This information can form the basis of machine learning to improve ad selection.

Image restoration

Deep learning has been successfully applied to

Deep Image Prior
, which trains on the image that needs restoration.

Financial fraud detection

Deep learning is being successfully applied to financial

fraud detection, tax evasion detection,[223] and anti-money laundering.[224]

Materials science

In November 2023, researchers at Google DeepMind and Lawrence Berkeley National Laboratory announced that they had developed an AI system known as GNoME. This system has contributed to materials science by discovering over 2 million new materials within a relatively short timeframe. GNoME employs deep learning techniques to efficiently explore potential material structures, achieving a significant increase in the identification of stable inorganic crystal structures. The system's predictions were validated through autonomous robotic experiments, demonstrating a noteworthy success rate of 71%. The data of newly discovered materials is publicly available through the Materials Project database, offering researchers the opportunity to identify materials with desired properties for various applications. This development has implications for the future of scientific discovery and the integration of AI in material science research, potentially expediting material innovation and reducing costs in product development. The use of AI and deep learning suggests the possibility of minimizing or eliminating manual lab experiments and allowing scientists to focus more on the design and analysis of unique compounds.[225][226][227]

Military

The United States Department of Defense applied deep learning to train robots in new tasks through observation.[228]

Partial differential equations

Physics informed neural networks have been used to solve partial differential equations in both forward and inverse problems in a data driven manner.[229] One example is the reconstructing fluid flow governed by the Navier-Stokes equations. Using physics informed neural networks does not require the often expensive mesh generation that conventional CFD methods relies on.[230][231]

Image reconstruction

Image reconstruction is the reconstruction of the underlying images from the image-related measurements. Several works showed the better and superior performance of the deep learning methods compared to analytical methods for various applications, e.g., spectral imaging [232] and ultrasound imaging.[233]

Epigenetic clock

Main article: Epigenetic clock

An epigenetic clock is a biochemical test that can be used to measure age. Galkin et al. used deep neural networks to train an epigenetic aging clock of unprecedented accuracy using >6,000 blood samples.[234] The clock uses information from 1000 CpG sites and predicts people with certain conditions older than healthy controls: IBD, frontotemporal dementia, ovarian cancer, obesity. The aging clock was planned to be released for public use in 2021 by an Insilico Medicine spinoff company Deep Longevity.

Relation to human cognitive and brain development

Deep learning is closely related to a class of theories of

cognitive neuroscientists in the early 1990s.[235][236][237][238] These developmental theories were instantiated in computational models, making them predecessors of deep learning systems. These developmental models share the property that various proposed learning dynamics in the brain (e.g., a wave of nerve growth factor) support the self-organization somewhat analogous to the neural networks utilized in deep learning models. Like the neocortex, neural networks employ a hierarchy of layered filters in which each layer considers information from a prior layer (or the operating environment), and then passes its output (and possibly the original input), to other layers. This process yields a self-organizing stack of transducers, well-tuned to their operating environment. A 1995 description stated, "...the infant's brain seems to organize itself under the influence of waves of so-called trophic-factors ... different regions of the brain become connected sequentially, with one layer of tissue maturing before another and so on until the whole brain is mature".[239]

A variety of approaches have been used to investigate the plausibility of deep learning models from a neurobiological perspective. On the one hand, several variants of the backpropagation algorithm have been proposed in order to increase its processing realism.[240][241] Other researchers have argued that unsupervised forms of deep learning, such as those based on hierarchical generative models and deep belief networks, may be closer to biological reality.[242][243] In this respect, generative neural network models have been related to neurobiological evidence about sampling-based processing in the cerebral cortex.[244]

Although a systematic comparison between the human brain organization and the neuronal encoding in deep networks has not yet been established, several analogies have been reported. For example, the computations performed by deep learning units could be similar to those of actual neurons[245] and neural populations.[246] Similarly, the representations developed by deep learning models are similar to those measured in the primate visual system[247] both at the single-unit[248] and at the population[249] levels.

Commercial activity

Facebook's AI lab performs tasks such as automatically tagging uploaded pictures with the names of the people in them.[250]

Google's

DeepMind Technologies developed a system capable of learning how to play Atari video games using only pixels as data input. In 2015 they demonstrated their AlphaGo system, which learned the game of Go well enough to beat a professional Go player.[251][252][253] Google Translate
uses a neural network to translate between more than 100 languages.

In 2017, Covariant.ai was launched, which focuses on integrating deep learning into factories.[254]

As of 2008,

U.S. Army Research Laboratory (ARL) and UT researchers. Deep TAMER used deep learning to provide a robot with the ability to learn new tasks through observation.[228] Using Deep TAMER, a robot learned a task with a human trainer, watching video streams or observing a human perform a task in-person. The robot later practiced the task with the help of some coaching from the trainer, who provided feedback such as "good job" and "bad job".[256]

Criticism and comment

Deep learning has attracted both criticism and comment, in some cases from outside the field of computer science.

Theory

See also: Explainable artificial intelligence

A main criticism concerns the lack of theory surrounding some methods.[257] Learning in the most common deep architectures is implemented using well-understood gradient descent. However, the theory surrounding other algorithms, such as contrastive divergence is less clear.[citation needed] (e.g., Does it converge? If so, how fast? What is it approximating?) Deep learning methods are often looked at as a black box, with most confirmations done empirically, rather than theoretically.[258]

Others point out that deep learning should be looked at as a step towards realizing strong AI, not as an all-encompassing solution. Despite the power of deep learning methods, they still lack much of the functionality needed to realize this goal entirely. Research psychologist Gary Marcus noted:

Realistically, deep learning is only part of the larger challenge of building intelligent machines. Such techniques lack ways of representing

Watson (...) use techniques like deep learning as just one element in a very complicated ensemble of techniques, ranging from the statistical technique of Bayesian inference to deductive reasoning.[259]

In further reference to the idea that artistic sensitivity might be inherent in relatively low levels of the cognitive hierarchy, a published series of graphic representations of the internal states of deep (20-30 layers) neural networks attempting to discern within essentially random data the images on which they were trained[260] demonstrate a visual appeal: the original research notice received well over 1,000 comments, and was the subject of what was for a time the most frequently accessed article on The Guardian's[261] website.

Errors

Some deep learning architectures display problematic behaviors,[262] such as confidently classifying unrecognizable images as belonging to a familiar category of ordinary images (2014)[263] and misclassifying minuscule perturbations of correctly classified images (2013).[264] Goertzel hypothesized that these behaviors are due to limitations in their internal representations and that these limitations would inhibit integration into heterogeneous multi-component artificial general intelligence (AGI) architectures.[262] These issues may possibly be addressed by deep learning architectures that internally form states homologous to image-grammar[265] decompositions of observed entities and events.[262] Learning a grammar (visual or linguistic) from training data would be equivalent to restricting the system to commonsense reasoning that operates on concepts in terms of grammatical production rules and is a basic goal of both human language acquisition[266] and artificial intelligence (AI).[267]

Cyber threat

As deep learning moves from the lab into the world, research and experience show that artificial neural networks are vulnerable to hacks and deception.[268] By identifying patterns that these systems use to function, attackers can modify inputs to ANNs in such a way that the ANN finds a match that human observers would not recognize. For example, an attacker can make subtle changes to an image such that the ANN finds a match even though the image looks to a human nothing like the search target. Such manipulation is termed an "adversarial attack".[269]

In 2016 researchers used one ANN to doctor images in trial and error fashion, identify another's focal points, and thereby generate images that deceived it. The modified images looked no different to human eyes. Another group showed that printouts of doctored images then photographed successfully tricked an image classification system.[270] One defense is reverse image search, in which a possible fake image is submitted to a site such as TinEye that can then find other instances of it. A refinement is to search using only parts of the image, to identify images from which that piece may have been taken.[271]

Another group showed that certain psychedelic spectacles could fool a facial recognition system into thinking ordinary people were celebrities, potentially allowing one person to impersonate another. In 2017 researchers added stickers to stop signs and caused an ANN to misclassify them.[270]

ANNs can however be further trained to detect attempts at deception, potentially leading attackers and defenders into an arms race similar to the kind that already defines the malware defense industry. ANNs have been trained to defeat ANN-based anti-malware software by repeatedly attacking a defense with malware that was continually altered by a genetic algorithm until it tricked the anti-malware while retaining its ability to damage the target.[270]

In 2016, another group demonstrated that certain sounds could make the Google Now voice command system open a particular web address, and hypothesized that this could "serve as a stepping stone for further attacks (e.g., opening a web page hosting drive-by malware)".[270]

In "data poisoning", false data is continually smuggled into a machine learning system's training set to prevent it from achieving mastery.[270]

Data collection ethics

Most Deep Learning systems rely on training and verification data that is generated and/or annotated by humans.

information mining (e.g. by leveraging quantified-self devices such as activity trackers) and (5) clickwork.[273]

Mühlhoff argues that in most commercial end-user applications of Deep Learning such as Facebook's face recognition system, the need for training data does not stop once an ANN is trained. Rather, there is a continued demand for human-generated verification data to constantly calibrate and update the ANN. For this purpose, Facebook introduced the feature that once a user is automatically recognized in an image, they receive a notification. They can choose whether or not they like to be publicly labeled on the image, or tell Facebook that it is not them in the picture.[274] This user interface is a mechanism to generate "a constant stream of verification data"[273] to further train the network in real-time. As Mühlhoff argues, the involvement of human users to generate training and verification data is so typical for most commercial end-user applications of Deep Learning that such systems may be referred to as "human-aided artificial intelligence".[273]

See also

References

  1. S2CID 220523562
    .
  2. S2CID 3074096
    .
  3. ^
    S2CID 2161592
    .
  4. ^ a b Krizhevsky, Alex; Sutskever, Ilya; Hinton, Geoffrey (2012). "ImageNet Classification with Deep Convolutional Neural Networks" (PDF). NIPS 2012: Neural Information Processing Systems, Lake Tahoe, Nevada. Archived (PDF) from the original on 2017-01-10. Retrieved 2017-05-24.
  5. ^ "Google's AlphaGo AI wins three-match series against the world's best Go player". TechCrunch. 25 May 2017. Archived from the original on 17 June 2018. Retrieved 17 June 2018.
  6. S2CID 1994856
    .
  7. arXiv:1502.04156 [cs.LG
    ].
  8. ^ "Study urges caution when comparing neural networks to the brain". MIT News | Massachusetts Institute of Technology. 2022-11-02. Retrieved 2023-12-06.
  9. ^
    doi:10.1561/2000000039. Archived
    (PDF) from the original on 2016-03-14. Retrieved 2014-10-18.
  10. S2CID 51971897
    .
  11. ^
    S2CID 207178999. Archived from the original
    (PDF) on 4 March 2016. Retrieved 3 September 2015.
  12. ^
    S2CID 393948
    .
  13. S2CID 3074096
    .
  14. ^
    S2CID 11715509
    .
  15. ISBN 978-1-5225-8218-2
    .
  16. ^ Bengio, Yoshua; Lamblin, Pascal; Popovici, Dan; Larochelle, Hugo (2007). Greedy layer-wise training of deep networks (PDF). Advances in neural information processing systems. pp. 153–160. Archived (PDF) from the original on 2019-10-20. Retrieved 2019-10-06.
  17. ^
    doi:10.4249/scholarpedia.5947
    .
  18. ISSN 2076-3417
    .
  19. ^
    S2CID 3958369. Archived from the original
    (PDF) on 10 October 2015.
  20. ^
    S2CID 7343126
    .
  21. ^
    ISBN 978-0-13-273350-2
    .
  22. ^
    ISBN 978-0-262-08239-6
    .
  23. ^ a b Lu, Z., Pu, H., Wang, F., Hu, Z., & Wang, L. (2017). The Expressive Power of Neural Networks: A View from the Width Archived 2019-02-13 at the Wayback Machine. Neural Information Processing Systems, 6231-6239.
  24. PMID 28743932
    .
  25. ^
    ISBN 978-0-262-01802-9
    .
  26. ^
    doi:10.1109/TSSC.1969.300225
    .
  27. S2CID 12149203
    .
  28. ISBN 978-0-387-31073-2. Archived
    (PDF) from the original on 2017-01-11. Retrieved 2017-08-06.
  29. doi:10.1103/RevModPhys.39.883
    .
  30. ^ Amari, Shun-Ichi (1972). "Learning patterns and pattern sequences by self-organizing nets of threshold elements". IEEE Transactions. C (21): 1197–1206.
  31. ^
    arXiv:2212.11279 [cs.NE
    ].
  32. PMID 6953413
    .
  33. S2CID 216043128
    . Retrieved 31 May 2021.
  34. ^ a b c Rosenblatt, Frank (1962). Principles of Neurodynamics. Spartan, New York.
  35. ^ a b Rina Dechter (1986). Learning while searching in constraint-satisfaction problems. University of California, Computer Science Department, Cognitive Systems Laboratory.Online Archived 2016-04-19 at the Wayback Machine
  36. S2CID 235081987
    .
  37. ISBN 978-0-444-00020-0
    .
  38. ^
    doi:10.1109/TSMC.1971.4308320. Archived
    (PDF) from the original on 2017-08-29. Retrieved 2019-11-05.
  39. doi:10.1214/aoms/1177729586
    .
  40. ^ Amari, Shun'ichi (1967). "A theory of adaptive pattern classifier". IEEE Transactions. EC (16): 279–307.
  41. ^ Matthew Brand (1988) Machine and Brain Learning. University of Chicago Tutorial Studies Bachelor's Thesis, 1988. Reported at the Summer Linguistics Institute, Stanford University, 1987
  42. ^ Linnainmaa, Seppo (1970). The representation of the cumulative rounding error of an algorithm as a Taylor expansion of the local rounding errors (Masters) (in Finnish). University of Helsinki. pp. 6–7.
  43. S2CID 122357351
    .
  44. ^ Griewank, Andreas (2012). "Who Invented the Reverse Mode of Differentiation?" (PDF). Documenta Mathematica (Extra Volume ISMP): 389–400. Archived from the original (PDF) on 21 July 2017. Retrieved 11 June 2017.
  45. ISBN 9780598818461
    .
  46. doi:10.2514/8.5282
    .
  47. ^ Werbos, Paul (1982). "Applications of advances in nonlinear sensitivity analysis". System modeling and optimization. Springer. pp. 762–770.
  48. ^ Werbos, P. (1974). "Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences". Harvard University. Retrieved 12 June 2017.
  49. ^ Rumelhart, David E., Geoffrey E. Hinton, and R. J. Williams. "Learning Internal Representations by Error Propagation". David E. Rumelhart, James L. McClelland, and the PDP research group. (editors), Parallel distributed processing: Explorations in the microstructure of cognition, Volume 1: Foundation. MIT Press, 1986.
  50. S2CID 206775608
    .
  51. arXiv:1710.05941 [cs.NE
    ].
  52. ISBN 978-0-7923-7824-2
    . Retrieved 27 December 2023.
  53. ^ Co-evolving recurrent neurons learn deep memory POMDPs. Proc. GECCO, Washington, D. C., pp. 1795–1802, ACM Press, New York, NY, USA, 2005.
  54. ^ a b Zhang, Wei (1988). "Shift-invariant pattern recognition neural network and its optical architecture". Proceedings of Annual Conference of the Japan Society of Applied Physics.
  55. PMID 20577468
    .
  56. ^ a b LeCun et al., "Backpropagation Applied to Handwritten Zip Code Recognition", Neural Computation, 1, pp. 541–551, 1989.
  57. PMID 20706526
    .
  58. PMID 8058017
    .
  59. S2CID 14542261
    . Retrieved October 7, 2016.
  60. ^
    S2CID 18271205.[permanent dead link
    ]
  61. ^ Schmidhuber, Jürgen (1993). Habilitation Thesis (PDF) (in German). Archived from the original (PDF) on 26 June 2021.
  62. S2CID 16683347
    .
  63. ^ a b Schlag, Imanol; Irie, Kazuki; Schmidhuber, Jürgen (2021). "Linear Transformers Are Secretly Fast Weight Programmers". ICML 2021. Springer. pp. 9355–9366.
  64. arXiv:2009.14794 [cs.CL
    ].
  65. ^ Schmidhuber, Jürgen (1993). "Reducing the ratio between learning complexity and number of time-varying variables in fully recurrent nets". ICANN 1993. Springer. pp. 460–463.
  66. arXiv:1706.03762 [cs.CL
    ].
  67. S2CID 208117506
    .
  68. ^ He, Cheng (31 December 2021). "Transformer in CV". Transformer in CV. Towards Data Science.
  69. ^ Schmidhuber, Jürgen (1991). "A possibility for implementing curiosity and boredom in model-building neural controllers". Proc. SAB'1991. MIT Press/Bradford Books. pp. 222–227.
  70. S2CID 234198
    .
  71. S2CID 216056336
    .
  72. ^ Goodfellow, Ian; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua (2014). Generative Adversarial Networks (PDF). Proceedings of the International Conference on Neural Information Processing Systems (NIPS 2014). pp. 2672–2680. Archived (PDF) from the original on 22 November 2019. Retrieved 20 August 2019.
  73. ^ "Prepare, Don't Panic: Synthetic Media and Deepfakes". witness.org. Archived from the original on 2 December 2020. Retrieved 25 November 2020.
  74. ^ "GAN 2.0: NVIDIA's Hyperrealistic Face Generator". SyncedReview.com. December 14, 2018. Retrieved October 3, 2019.
  75. arXiv:1710.10196 [cs.NE
    ].
  76. ^ a b c S. Hochreiter., "Untersuchungen zu dynamischen neuronalen Netzen". Archived 2015-03-06 at the Wayback Machine. Diploma thesis. Institut f. Informatik, Technische Univ. Munich. Advisor: J. Schmidhuber, 1991.
  77. ISBN 978-0-7803-5369-5
    .
  78. ^
    S2CID 1915014
    .
  79. ISBN 0-85296-721-7
    .
  80. arXiv:1505.00387 [cs.LG
    ].
  81. ^ Srivastava, Rupesh K; Greff, Klaus; Schmidhuber, Jürgen (2015). "Training Very Deep Networks". Advances in Neural Information Processing Systems. 28. Curran Associates, Inc.: 2377–2385.
  82. ISBN 978-1-4673-8851-1
    .
  83. doi:10.1016/0167-8655(94)90009-4
    .
  84. S2CID 871473
    .
  85. S2CID 1304548
    .
  86. ISSN 0218-0014
    .
  87. ISBN 9780780305328. Archived
    from the original on 2021-05-09. Retrieved 2017-06-12.
  88. S2CID 9563026. Archived
    (PDF) from the original on 2021-04-27. Retrieved 2019-09-24.
  89. S2CID 357467
    .
  90. ^ Bengio, Y. (1991). "Artificial Neural Networks and their Application to Speech/Sequence Recognition". McGill University Ph.D. thesis. Archived from the original on 2021-05-09. Retrieved 2017-06-12.
  91. doi:10.1016/0893-6080(94)90027-2
    .
  92. doi:10.1016/S0167-6393(99)00080-1
    .
  93. ^
    doi:10.1016/s0167-6393(99)00077-1
    .
  94. ^ "Acoustic Modeling with Deep Neural Networks Using Raw Time Signal for LVCSR (PDF Download Available)". ResearchGate. Archived from the original on 9 May 2021. Retrieved 14 June 2017.
  95. ^ a b Graves, Alex; Eck, Douglas; Beringer, Nicole; Schmidhuber, Jürgen (2003). "Biologically Plausible Speech Recognition with LSTM Neural Nets" (PDF). 1st Intl. Workshop on Biologically Inspired Approaches to Advanced Information Technology, Bio-ADIT 2004, Lausanne, Switzerland. pp. 175–184. Archived (PDF) from the original on 2021-05-09. Retrieved 2016-04-09.
  96. ^
    CiteSeerX 10.1.1.75.6306
    .
  97. ^ Santiago Fernandez, Alex Graves, and Jürgen Schmidhuber (2007). An application of recurrent neural networks to discriminative keyword spotting Archived 2018-11-18 at the Wayback Machine. Proceedings of ICANN (2), pp. 220–229.
  98. ^ a b c Sak, Haşim; Senior, Andrew; Rao, Kanishka; Beaufays, Françoise; Schalkwyk, Johan (September 2015). "Google voice search: faster and more accurate". Archived from the original on 2016-03-09. Retrieved 2016-04-09.
  99. ^ Yann LeCun (2016). Slides on Deep Learning Online Archived 2016-04-23 at the Wayback Machine
  100. S2CID 15066318. Archived
    from the original on 11 October 2013. Retrieved 12 June 2017.
  101. S2CID 2309950. Archived
    (PDF) from the original on 2015-12-23. Retrieved 2011-07-20.
  102. arXiv:1206.5533 [cs.LG
    ].
  103. ^ G. E. Hinton., "Learning multiple layers of representation". Archived 2018-05-22 at the Wayback Machine. Trends in Cognitive Sciences, 11, pp. 428–434, 2007.
  104. ^
    S2CID 206485943
    .
  105. ^ a b c Deng, L.; Hinton, G.; Kingsbury, B. (May 2013). "New types of deep neural network learning for speech recognition and related applications: An overview (ICASSP)" (PDF). Microsoft. Archived (PDF) from the original on 2017-09-26. Retrieved 27 December 2023.
  106. ^
    ISBN 978-1-4471-5779-3
    .
  107. ^ "Deng receives prestigious IEEE Technical Achievement Award - Microsoft Research". Microsoft Research. 3 December 2015. Archived from the original on 16 March 2018. Retrieved 16 March 2018.
  108. ^ a b Li, Deng (September 2014). "Keynote talk: 'Achievements and Challenges of Deep Learning - From Speech Analysis and Recognition To Language and Multimodal Processing'". Interspeech. Archived from the original on 2017-09-26. Retrieved 2017-06-12.
  109. ^ Yu, D.; Deng, L. (2010). "Roles of Pre-Training and Fine-Tuning in Context-Dependent DBN-HMMs for Real-World Speech Recognition". NIPS Workshop on Deep Learning and Unsupervised Feature Learning. Archived from the original on 2017-10-12. Retrieved 2017-06-14.
  110. S2CID 398770. Archived
    from the original on 2017-10-12. Retrieved 2017-06-14.
  111. ^ Deng, Li; Li, Jinyu; Huang, Jui-Ting; Yao, Kaisheng; Yu, Dong; Seide, Frank; Seltzer, Mike; Zweig, Geoff; He, Xiaodong (1 May 2013). "Recent Advances in Deep Learning for Speech Research at Microsoft". Microsoft Research. Archived from the original on 12 October 2017. Retrieved 14 June 2017.
  112. S2CID 231846518
    .
  113. ^ Sak, Hasim; Senior, Andrew; Beaufays, Francoise (2014). "Long Short-Term Memory recurrent neural network architectures for large scale acoustic modeling" (PDF). Archived from the original (PDF) on 24 April 2018.
  114. arXiv:1410.4281 [cs.CL
    ].
  115. ^ Zen, Heiga; Sak, Hasim (2015). "Unidirectional Long Short-Term Memory Recurrent Neural Network with Recurrent Output Layer for Low-Latency Speech Synthesis" (PDF). Google.com. ICASSP. pp. 4470–4474. Archived (PDF) from the original on 2021-05-09. Retrieved 2017-06-13.
  116. Venture Beat. 5 April 2016. Archived
    from the original on 25 November 2020. Retrieved 21 April 2017.
  117. ^ "From not working to neural networking". The Economist. Archived from the original on 2016-12-31. Retrieved 2017-08-26.
  118. ^
    doi:10.1016/j.patcog.2004.01.013
    .
  119. ^ "A Survey of Techniques for Optimizing Deep Learning on GPUs Archived 2021-05-09 at the Wayback Machine", S. Mittal and S. Vaishay, Journal of Systems Architecture, 2019
  120. ^ a b Chellapilla, Kumar; Puri, Sidd; Simard, Patrice (2006), High performance convolutional neural networks for document processing, archived from the original on 2020-05-18, retrieved 2021-02-14
  121. S2CID 1918673
    .
  122. S2CID 392458
    .
  123. arXiv:1703.09039 [cs.CV
    ].
  124. ^ Graves, Alex; and Schmidhuber, Jürgen; Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks, in Bengio, Yoshua; Schuurmans, Dale; Lafferty, John; Williams, Chris K. I.; and Culotta, Aron (eds.), Advances in Neural Information Processing Systems 22 (NIPS'22), December 7th–10th, 2009, Vancouver, BC, Neural Information Processing Systems (NIPS) Foundation, 2009, pp. 545–552
  125. ^ Google Research Blog. The neural networks behind Google Voice transcription. August 11, 2015. By Françoise Beaufays http://googleresearch.blogspot.co.at/2015/08/the-neural-networks-behind-google-voice.html
  126. doi:10.5591/978-1-57735-516-8/ijcai11-210. Archived
    (PDF) from the original on 2014-09-29. Retrieved 2017-06-13.
  127. ^ Ciresan, Dan; Giusti, Alessandro; Gambardella, Luca M.; Schmidhuber, Jürgen (2012). Pereira, F.; Burges, C. J. C.; Bottou, L.; Weinberger, K. Q. (eds.). Advances in Neural Information Processing Systems 25 (PDF). Curran Associates, Inc. pp. 2843–2851. Archived (PDF) from the original on 2017-08-09. Retrieved 2017-06-13.
  128. PMID 24579167
    .
  129. arXiv:1409.1556 [cs.CV
    ].
  130. arXiv:1411.4555 [cs.CV
    ]..
  131. arXiv:1411.4952 [cs.CV
    ]..
  132. arXiv:1411.2539 [cs.LG
    ]..
  133. ^ a b "Merck Molecular Activity Challenge". kaggle.com. Archived from the original on 2020-07-16. Retrieved 2020-07-16.
  134. ^ a b "Multi-task Neural Networks for QSAR Predictions | Data Science Association". www.datascienceassn.org. Archived from the original on 30 April 2017. Retrieved 14 June 2017.
  135. ^ a b "Toxicology in the 21st century Data Challenge"
  136. ^ a b "NCATS Announces Tox21 Data Challenge Winners". Archived from the original on 2015-09-08. Retrieved 2015-03-05.
  137. ^ a b "NCATS Announces Tox21 Data Challenge Winners". Archived from the original on 28 February 2015. Retrieved 5 March 2015.
  138. ^ "Why Deep Learning Is Suddenly Changing Your Life". Fortune. 2016. Archived from the original on 14 April 2018. Retrieved 13 April 2018.
  139. ISBN 978-1492671206.{{cite book}}: CS1 maint: multiple names: authors list (link
    )
  140. S2CID 515925
    .
  141. ^ A Guide to Deep Learning and Neural Networks, archived from the original on 2020-11-02, retrieved 2020-11-16
  142. ^ Szegedy, Christian; Toshev, Alexander; Erhan, Dumitru (2013). "Deep neural networks for object detection". Advances in Neural Information Processing Systems: 2553–2561. Archived from the original on 2017-06-29. Retrieved 2017-06-13.
  143. ^ Rolnick, David; Tegmark, Max (2018). "The power of deeper networks for expressing natural functions". International Conference on Learning Representations. ICLR 2018. Archived from the original on 2021-01-07. Retrieved 2021-01-05.
  144. ^ Hof, Robert D. "Is Artificial Intelligence Finally Coming into Its Own?". MIT Technology Review. Archived from the original on 31 March 2019. Retrieved 10 July 2018.
  145. ^
    S2CID 10192330. Archived
    from the original on 2020-01-26. Retrieved 2020-02-25.
  146. ^
    Bibcode:2014arXiv1409.3215S. Archived
    (PDF) from the original on 2021-05-09. Retrieved 2017-06-13.
  147. ^
    arXiv:1602.02410 [cs.CL
    ].
  148. ^
    arXiv:1512.00103 [cs.CL
    ].
  149. S2CID 17048224. Archived
    (PDF) from the original on 2017-05-16. Retrieved 2017-06-13.
  150. ^ a b "Learning Precise Timing with LSTM Recurrent Networks (PDF Download Available)". ResearchGate. Archived from the original on 9 May 2021. Retrieved 13 June 2017.
  151. S2CID 14542261
    .
  152. S2CID 13816461
    .
  153. S2CID 12485056
    .
  154. ^ Dahl, G.; et al. (2013). "Improving DNNs for LVCSR using rectified linear units and dropout" (PDF). ICASSP. Archived (PDF) from the original on 2017-08-12. Retrieved 2017-06-13.
  155. ^ "Data Augmentation - deeplearning.ai | Coursera". Coursera. Archived from the original on 1 December 2017. Retrieved 30 November 2017.
  156. ^ Hinton, G. E. (2010). "A Practical Guide to Training Restricted Boltzmann Machines". Tech. Rep. UTML TR 2010-003. Archived from the original on 2021-05-09. Retrieved 2017-06-13.
  157. S2CID 8869270. Archived
    from the original on 29 July 2020. Retrieved 5 March 2018.
  158. S2CID 14135321
    .
  159. ^ Ting Qin, et al. "A learning algorithm of CMAC based on RLS". Neural Processing Letters 19.1 (2004): 49-61.
  160. ^ Ting Qin, et al. "Continuous CMAC-QRLS and its systolic array". Archived 2018-11-18 at the Wayback Machine. Neural Processing Letters 22.1 (2005): 1-16.
  161. ^ Research, AI (23 October 2015). "Deep Neural Networks for Acoustic Modeling in Speech Recognition". airesearch.com. Archived from the original on 1 February 2016. Retrieved 23 October 2015.
  162. ^ "GPUs Continue to Dominate the AI Accelerator Market for Now". InformationWeek. December 2019. Archived from the original on 10 June 2020. Retrieved 11 June 2020.
  163. ^ Ray, Tiernan (2019). "AI is changing the entire nature of computation". ZDNet. Archived from the original on 25 May 2020. Retrieved 11 June 2020.
  164. ^ "AI and Compute". OpenAI. 16 May 2018. Archived from the original on 17 June 2020. Retrieved 11 June 2020.
  165. ^ "HUAWEI Reveals the Future of Mobile AI at IFA 2017 | HUAWEI Latest News | HUAWEI Global". consumer.huawei.com.
  166. doi:10.1145/3140659.3080246
    .
  167. ^ Woodie, Alex (2021-11-01). "Cerebras Hits the Accelerator for Deep Learning Workloads". Datanami. Retrieved 2022-08-03.
  168. ^ "Cerebras launches new AI supercomputing processor with 2.6 trillion transistors". VentureBeat. 2021-04-20. Retrieved 2022-08-03.
  169. PMID 33149289
    .
  170. ^
    S2CID 211010976
    .
  171. ISBN 1-58563-019-5
    . Retrieved 27 December 2023.
  172. doi:10.13140/RG.2.2.15418.90567
    .
  173. S2CID 206602362. Archived
    from the original on 2020-09-22. Retrieved 2018-04-20.
  174. S2CID 15641618
    .
  175. S2CID 217950236. Archived
    (PDF) from the original on 2020-09-24. Retrieved 2019-04-01.
  176. ^ McMillan, Robert (17 December 2014). "How Skype Used AI to Build Its Amazing New Language Translator | WIRED". Wired. Archived from the original on 8 June 2017. Retrieved 14 June 2017.
  177. arXiv:1412.5567 [cs.CL
    ].
  178. ^ "MNIST handwritten digit database, Yann LeCun, Corinna Cortes and Chris Burges". yann.lecun.com. Archived from the original on 2014-01-13. Retrieved 2014-01-28.
  179. PMID 22386783
    .
  180. arXiv:1404.3840 [cs.CV
    ].
  181. ^ Nvidia Demos a Car Computer Trained with "Deep Learning" (6 January 2015), David Talbot, MIT Technology Review
  182. ^
    doi:10.3390/arts6020005
    .
  183. ^
    doi:10.3390/arts6040018
    .
  184. arXiv:1402.3722 [cs.CL
    ].
  185. ^ a b Socher, Richard; Manning, Christopher. "Deep Learning for NLP" (PDF). Archived (PDF) from the original on 6 July 2014. Retrieved 26 October 2014.
  186. ^ Socher, Richard; Bauer, John; Manning, Christopher; Ng, Andrew (2013). "Parsing With Compositional Vector Grammars" (PDF). Proceedings of the ACL 2013 Conference. Archived (PDF) from the original on 2014-11-27. Retrieved 2014-09-03.
  187. ^ Socher, R.; Perelygin, A.; Wu, J.; Chuang, J.; Manning, C.D.; Ng, A.; Potts, C. (October 2013). "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank" (PDF). Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics. Archived (PDF) from the original on 28 December 2016. Retrieved 21 December 2023.
  188. ^ Shen, Yelong; He, Xiaodong; Gao, Jianfeng; Deng, Li; Mesnil, Gregoire (1 November 2014). "A Latent Semantic Model with Convolutional-Pooling Structure for Information Retrieval". Microsoft Research. Archived from the original on 27 October 2017. Retrieved 14 June 2017.
  189. ^ Huang, Po-Sen; He, Xiaodong; Gao, Jianfeng; Deng, Li; Acero, Alex; Heck, Larry (1 October 2013). "Learning Deep Structured Semantic Models for Web Search using Clickthrough Data". Microsoft Research. Archived from the original on 27 October 2017. Retrieved 14 June 2017.
  190. S2CID 1317136
    .
  191. ^ a b Gao, Jianfeng; He, Xiaodong; Yih, Scott Wen-tau; Deng, Li (1 June 2014). "Learning Continuous Phrase Representations for Translation Modeling". Microsoft Research. Archived from the original on 27 October 2017. Retrieved 14 June 2017.
  192. S2CID 40745740
    .
  193. PMID 36855134
    .
  194. ^ "Deep Learning for Natural Language Processing: Theory and Practice (CIKM2014 Tutorial) - Microsoft Research". Microsoft Research. Archived from the original on 13 March 2017. Retrieved 14 June 2017.
  195. ^ Turovsky, Barak (15 November 2016). "Found in translation: More accurate, fluent sentences in Google Translate". The Keyword Google Blog. Archived from the original on 7 April 2017. Retrieved 23 March 2017.
  196. ^ a b c d Schuster, Mike; Johnson, Melvin; Thorat, Nikhil (22 November 2016). "Zero-Shot Translation with Google's Multilingual Neural Machine Translation System". Google Research Blog. Archived from the original on 10 July 2017. Retrieved 23 March 2017.
  197. arXiv:1609.08144 [cs.CL
    ].
  198. ^ Metz, Cade (27 September 2016). "An Infusion of AI Makes Google Translate More Powerful Than Ever". Wired. Archived from the original on 8 November 2020. Retrieved 12 October 2017.
  199. ^ a b Boitet, Christian; Blanchon, Hervé; Seligman, Mark; Bellynck, Valérie (2010). "MT on and for the Web" (PDF). Archived from the original (PDF) on 29 March 2017. Retrieved 1 December 2016.
  200. S2CID 20246434
    .
  201. PMID 25582842
    .
  202. arXiv:1510.02855 [cs.LG
    ].
  203. ^ a b "Toronto startup has a faster way to discover effective medicines". The Globe and Mail. Archived from the original on 20 October 2015. Retrieved 9 November 2015.
  204. ^ "Startup Harnesses Supercomputers to Seek Cures". KQED Future of You. 27 May 2015. Archived from the original on 24 December 2015. Retrieved 9 November 2015.
  205. arXiv:1704.01212 [cs.LG
    ].
  206. S2CID 201716327
    .
  207. ^ Gregory, Barber. "A Molecule Designed By AI Exhibits 'Druglike' Qualities". Wired. Archived from the original on 2020-04-30. Retrieved 2019-09-05.
  208. arXiv:1504.01840 [cs.LG
    ].
  209. ^ van den Oord, Aaron; Dieleman, Sander; Schrauwen, Benjamin (2013). Burges, C. J. C.; Bottou, L.; Welling, M.; Ghahramani, Z.; Weinberger, K. Q. (eds.). Advances in Neural Information Processing Systems 26 (PDF). Curran Associates, Inc. pp. 2643–2651. Archived (PDF) from the original on 2017-05-16. Retrieved 2017-06-14.
  210. PMID 31127715
    .
  211. ^ Elkahky, Ali Mamdouh; Song, Yang; He, Xiaodong (1 May 2015). "A Multi-View Deep Learning Approach for Cross Domain User Modeling in Recommendation Systems". Microsoft Research. Archived from the original on 25 January 2018. Retrieved 14 June 2017.
  212. S2CID 207217210. Archived
    from the original on 9 May 2021. Retrieved 23 November 2015.
  213. S2CID 3821594
    .
  214. PMID 27521897
    .
  215. ^
    S2CID 229339809
    .
  216. S2CID 2088679
    .
  217. S2CID 4728736. Archived
    from the original on 2021-05-09. Retrieved 2019-11-12.
  218. S2CID 220733699
    .
  219. PMID 35406591
    .
  220. S2CID 35350962
    .
  221. ^ "Colorizing and Restoring Old Images with Deep Learning". FloydHub Blog. 13 November 2018. Archived from the original on 11 October 2019. Retrieved 11 October 2019.
  222. ^ Schmidt, Uwe; Roth, Stefan. Shrinkage Fields for Effective Image Restoration (PDF). Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on. Archived (PDF) from the original on 2018-01-02. Retrieved 2018-01-01.
  223. S2CID 204092079
    .
  224. ^ Czech, Tomasz (28 June 2018). "Deep learning: the next frontier for money laundering detection". Global Banking and Finance Review. Archived from the original on 2018-11-16. Retrieved 2018-07-15.
  225. ^ Nuñez, Michael (2023-11-29). "Google DeepMind's materials AI has already discovered 2.2 million new crystals". VentureBeat. Retrieved 2023-12-19.
  226. PMID 38030720
    .
  227. S2CID 265503872
    .
  228. ^ a b c "Army researchers develop new algorithms to train robots". EurekAlert!. Archived from the original on 28 August 2018. Retrieved 29 August 2018.
  229. S2CID 57379996
    .
  230. S2CID 212755458
    .
  231. PMID 32001523
    .
  232. S2CID 235340737
    .
  233. S2CID 220055785
    .
  234. doi:10.14336/AD
    .
  235. S2CID 1119517
    .
  236. ISBN 978-0-262-55030-7
    .
  237. PMID 12662587
    .
  238. S2CID 5818342
    .
  239. ^ S. Blakeslee, "In brain's early growth, timetable may be critical", The New York Times, Science Section, pp. B5–B6, 1995.
  240. PMID 1903542
    .
  241. S2CID 2376781
    .
  242. S2CID 9868901
    .
  243. S2CID 24504018
    .
  244. S2CID 7504633
    .
  245. S2CID 14663106
    .
  246. S2CID 16560320
    .
  247. S2CID 16970545
    .
  248. S2CID 39281431
    .
  249. PMID 26157000
    .
  250. ^ Metz, C. (12 December 2013). "Facebook's 'Deep Learning' Guru Reveals the Future of AI". Wired. Archived from the original on 28 March 2014. Retrieved 26 August 2017.
  251. S2CID 4460235
    .
  252. S2CID 515925.Closed access icon
  253. ^ "A Google DeepMind Algorithm Uses Deep Learning and More to Master the Game of Go | MIT Technology Review". MIT Technology Review. Archived from the original on 1 February 2016. Retrieved 30 January 2016.
  254. ^ Metz, Cade (6 November 2017). "A.I. Researchers Leave Elon Musk Lab to Begin Robotics Start-Up". The New York Times. Archived from the original on 7 July 2019. Retrieved 5 July 2019.
  255. S2CID 5613334
    .
  256. ^ "Talk to the Algorithms: AI Becomes a Faster Learner". governmentciomedia.com. 16 May 2018. Archived from the original on 28 August 2018. Retrieved 29 August 2018.
  257. ^ Marcus, Gary (14 January 2018). "In defense of skepticism about deep learning". Gary Marcus. Archived from the original on 12 October 2018. Retrieved 11 October 2018.
  258. ^ Knight, Will (14 March 2017). "DARPA is funding projects that will try to open up AI's black boxes". MIT Technology Review. Archived from the original on 4 November 2019. Retrieved 2 November 2017.
  259. ^ Marcus, Gary (November 25, 2012). "Is "Deep Learning" a Revolution in Artificial Intelligence?". The New Yorker. Archived from the original on 2009-11-27. Retrieved 2017-06-14.
  260. ^ Alexander Mordvintsev; Christopher Olah; Mike Tyka (17 June 2015). "Inceptionism: Going Deeper into Neural Networks". Google Research Blog. Archived from the original on 3 July 2015. Retrieved 20 June 2015.
  261. ^ Alex Hern (18 June 2015). "Yes, androids do dream of electric sheep". The Guardian. Archived from the original on 19 June 2015. Retrieved 20 June 2015.
  262. ^ a b c Goertzel, Ben (2015). "Are there Deep Reasons Underlying the Pathologies of Today's Deep Learning Algorithms?" (PDF). Archived (PDF) from the original on 2015-05-13. Retrieved 2015-05-10.
  263. arXiv:1412.1897 [cs.CV
    ].
  264. arXiv:1312.6199 [cs.CV
    ].
  265. doi:10.1561/0600000018
    .
  266. ^ Miller, G. A., and N. Chomsky. "Pattern conception". Paper for Conference on pattern detection, University of Michigan. 1957.
  267. ^ Eisner, Jason. "Deep Learning of Recursive Structure: Grammar Induction". Archived from the original on 2017-12-30. Retrieved 2015-05-10.
  268. ^ "Hackers Have Already Started to Weaponize Artificial Intelligence". Gizmodo. 11 September 2017. Archived from the original on 11 October 2019. Retrieved 11 October 2019.
  269. ^ "How hackers can force AI to make dumb mistakes". The Daily Dot. 18 June 2018. Archived from the original on 11 October 2019. Retrieved 11 October 2019.
  270. ^ a b c d e "AI Is Easy to Fool—Why That Needs to Change". Singularity Hub. 10 October 2017. Archived from the original on 11 October 2017. Retrieved 11 October 2017.
  271. doi:10.1038/nature.2017.22784. Archived
    from the original on 2017-10-10. Retrieved 2017-10-11.
  272. ^ Tubaro, Paola (2020). "Whose intelligence is artificial intelligence?". Global Dialogue: 38–39.
  273. ^
    S2CID 209363848
    .
  274. ISSN 1059-1028. Archived
    from the original on 10 August 2019. Retrieved 22 November 2019.

Further reading

Retrieved from "https://en.wikipedia.org/w/index.php?title=Deep_learning&oldid=1219088966"