This course teaches advanced technologies of artificial intelligence. This course consists of two parts. The topics of the first part include evolutionary computation. In the second part, students will learn reinforcement learning. Both techniques have a feature that they can find good solutions or strategies by trial and error. The aims of this course is to enable students 1) to acquire knowledge on evolutionary computation and reinforcement learning, and 2) to apply the knowledge to solve real-world problems.
By the end of this course, students will learn the following:
1) Evolutionary computation techniques and how to apply them to real-world problems.
2) Reinforcement learning techniques and how to apply them to real-world problems.
evolutionary computation,black-box optimization, multiobjective optimization, reinforcement learning, value-based methods, policy-based methods, deep reinforcement learning
|✔ Specialist skills||Intercultural skills||Communication skills||Critical thinking skills||Practical and/or problem-solving skills|
Every class consists of a lecture using the slides and the exercise.
|Course schedule||Required learning|
|Class 1||Introduction to evolutionary computation||Understand the aim of the course and foundation of evolutionary computation.|
|Class 2||Evolutionary computation for function optimization: Genetic algorithms||Understand function optimization and genetic algorithms.|
|Class 3||Evolutionary computation for function optimization : Evolution strategies||Understand evolution strategies.|
|Class 4||Evolutionary computation for combinatorial optimization:Genetic algorithms||Understand genetic algorithms for combinatorial optimization.|
|Class 5||Evolutionary computation for discrete optimization : Estimation of distribution algorithms||Understand estimation of distribution algorithms for black-box discrete function optmization.|
|Class 6||Evolutionary computation for Globally multimodal optimization||Understand global multimodality and evolutionary computation for globally multimodal optmization.|
|Class 7||Evolutionary computation for multiobjective optimization||Understand multiobjective optimization and evolutionary multiobjective optimization.|
|Class 8||Introduction to reinforcement learning||Understand foundation of reinforcement learning.|
|Class 9||Deep neural networks||Understand deep reinforcement neural networks.|
|Class 10||Deep Q-Network (DQN)||Understand the Deep-Q Network (DQN).|
|Class 11||Improvement of DQN||Understand the improved variants of DQN.|
|Class 12||Policy gradient and actor-critic methods||Understand REINFORCE, A2C, and A3C.|
|Class 13||Trust Region Policy Optimization and Proximal Policy Optimization||Understand Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO)|
|Class 14||Deep Deterministic Policy Gradient, Twin Delayed Deterministic Policy Gradient, and Soft Actor-Critic||Understand Deep Deterministic Policy Gradient (DDPG), Twin Delayed Deterministic Policy Gradient (TD3), and Soft Actor-Critic (SAC).|
To enhance effective learning, students are encouraged to spend approximately 100 minutes preparing for class and another 100 minutes reviewing class content afterwards (including assignments) for each class.
They should do so by referring to textbooks and other course material.
No textbook is set. Materials are distributed before each lesson.
Artificial Intelligence - A Modern Approach (Third Edition, Prentice Hall), and so on.
Students’ scores are based on assignment.
It is desiarble that studens have programming experience in Java and Python.