Dimitri Bertsekas

Dimitri Panteli Bertsekas (born 1942,

In the late 1990s Bertsekas developed a strong interest in digital photography. His photographs have been exhibited on several occasions at MIT.^[11]

Awards and honors

Bertsekas was elevated to the grade of

• Dynamic Programming and Optimal Control (1996)
• Data Networks (1989, co-authored with Robert G. Gallager)
• Nonlinear Programming (1996)
• Introduction to Probability (2003, co-authored with John N. Tsitsiklis)
• A Course in Reinforcement Learning (2023)

Monographs

• "Stochastic Optimal Control: The Discrete-Time Case" (1978, co-authored with S. E. Shreve), a mathematically complex work, establishing the measure-theoretic foundations of dynamic programming and stochastic control.
• "Constrained Optimization and Lagrange Multiplier Methods" (1982), the first monograph that addressed comprehensively the algorithmic convergence issues around
  augmented Lagrangian and sequential quadratic programming
  methods.
• "Parallel and Distributed Computation: Numerical Methods" (1989, co-authored with John N. Tsitsiklis), which among others established the fundamental theoretical structures for the analysis of distributed asynchronous algorithms.
• "Linear Network Optimization" (1991) and "Network Optimization: Continuous and Discrete Models" (1998), which among others discuss comprehensively the class of auction algorithms for assignment and network flow optimization, developed by Bertsekas over a period of 20 years starting in 1979.
• "Neuro-Dynamic Programming" (1996, co-authored with Tsitsiklis), which laid the theoretical foundations for suboptimal approximations of highly complex sequential decision-making problems.
• "Convex Analysis and Optimization" (2003, co-authored with A. Nedic and A. Ozdaglar) and "Convex Optimization Theory" (2009), which provided a new line of development for optimization duality theory, a new connection between the theory of
  subgradient methods
  .
• "Abstract Dynamic Programming" (2013), which aims at a unified development of the core theory and algorithms of total cost sequential decision problems, based on the strong connections of the subject with fixed point theory. A 3rd edition of this monograph, which extends the framework for applications to sequential zero-sum games and minimax problems, was published in 2022.
• "Reinforcement Learning and Optimal Control" (2019), which aims to explore the common boundary between dynamic programming/optimal control and artificial intelligence, and to form a bridge that is accessible by workers with background in either field.
• "Rollout, Policy Iteration, and Distributed Reinforcement Learning" (2020), which focuses on the fundamental idea of policy iteration, its one iteration counterpart, rollout, and their distributed and multiagent implementations. Some of these methods have been the backbones for high-profile successes in games such as chess, Go, and backgammon.^[22][23][24]
• “Lessons from AlphaZero for Optimal, Model Predictive, and Adaptive Control" (2022), which introduces a new conceptual framework for reinforcement learning, based on off-line training and on-line play algorithms, which are designed independently of each other but operate in synergy through the powerful mechanism of Newton's method.

See also

• Neural network
• Reinforcement learning
• Data network
• List of convexity topics

References