Lecture 12 2024; Off-line training with neural nets for approximate VI and PI. Aggregation

Автор: Dimitri Bertsekas

Загружено: 2024-04-06

Просмотров: 398

Описание:

Slides, class notes, and related textbook material at http://web.mit.edu/dimitrib/www/RLboo... A review of neural nets, approximation architectures, and off-line training. Approximate (fitted) value iteration, advantages of Q-learning, use of baselines, differential training, advantage updating. Implementation issues in approximate policy iteration: exploration, policy oscillations, robustness in the face of changing system parameters and on-line replanning. Aggregation architectures. A simple form of aggregation: representative states. Aggregation with representative features.

Lecture 12 2024; Off-line training with neural nets for approximate VI and PI. Aggregation

Доступные форматы для скачивания:

Скачать видео mp4

Информация по загрузке:

Скачать аудио mp3

Похожие видео

Lecture 11, 2024: On-line training, neural networks, and other approximation architectures

Lecture 11, 2024: On-line training, neural networks, and other approximation architectures

2: Training Deep NNs (cont.); Introduction to Keras/Tensorflow; Application to Tabular Data

2: Training Deep NNs (cont.); Introduction to Keras/Tensorflow; Application to Tabular Data

Reinforcement Learning Course at ASU

Reinforcement Learning Course at ASU

Computer chess with model predictive control and reinforcement learning

Computer chess with model predictive control and reinforcement learning

VAE in 10 Minutes 🧠✨ | The Clearest Variational Autoencoder Explanation

VAE in 10 Minutes 🧠✨ | The Clearest Variational Autoencoder Explanation

Bertsekas - Dynamic Programming

Bertsekas - Dynamic Programming

MIT 14.02 Principles of Macroeconomics, Spring 2023

MIT 14.02 Principles of Macroeconomics, Spring 2023

MIT 6.S191: Recurrent Neural Networks, Transformers, and Attention

MIT 6.S191: Recurrent Neural Networks, Transformers, and Attention

Reinforcement Learning, Model Predictive Control, and the Newton Step for Solving Bellman's Equation

Reinforcement Learning, Model Predictive Control, and the Newton Step for Solving Bellman's Equation

Код работает в 100 раз медленнее из-за ложного разделения ресурсов.

Код работает в 100 раз медленнее из-за ложного разделения ресурсов.

Abstract Dynamic Programming, Reinforcement Learning, Newton's Method, and Gradient Optimization

Abstract Dynamic Programming, Reinforcement Learning, Newton's Method, and Gradient Optimization

Я в опасности

Lecture 4, 2025, POMDP, Systems with Changing Parameters, Adaptive Control, Model Predictive Control

Lecture 4, 2025, POMDP, Systems with Changing Parameters, Adaptive Control, Model Predictive Control

Plenary lecture at IFAC Nonlinear MPC, 2024; Model Predictive Control and Reinforcement Learning

Plenary lecture at IFAC Nonlinear MPC, 2024; Model Predictive Control and Reinforcement Learning

Lec 1 | MIT 18.01 Single Variable Calculus, Fall 2007

Lec 1 | MIT 18.01 Single Variable Calculus, Fall 2007

Lecture 10, 2025; Aggregation Methods for Off-Line Training, Applications to POMDP and Cybersecurity

Lecture 10, 2025; Aggregation Methods for Off-Line Training, Applications to POMDP and Cybersecurity

Lecture 1, 2024, course overview: RL and DP, AlphaZero, discrete and continuous applications

Lecture 1, 2024, course overview: RL and DP, AlphaZero, discrete and continuous applications

New Directions in RL: TD(lambda), aggregation, seminorm projections, free-form sampling (from 2014)

New Directions in RL: TD(lambda), aggregation, seminorm projections, free-form sampling (from 2014)

Lecture 12, 2025; Training of cost functions, approximation in policy space, policy gradient methods

Lecture 12, 2025; Training of cost functions, approximation in policy space, policy gradient methods

Lecture 13 2024: Approximate LP. Approximation in policy space, policy gradient methods. Epilogue

Lecture 13 2024: Approximate LP. Approximation in policy space, policy gradient methods. Epilogue