Hyper-parameter optimization

目前主要方法有:

• 网格搜索
• 随机搜索
• 贝叶斯优化（效果最好）
• 遗传算法

Reinforcement learning

Learning to optimize

Learning to learn by gradient descent by gradient descent

Using Deep Q-Learning to Control Optimization Hyperparameters

Meta-learning

Optimization as a model for few-shot learning

A bridge between hyperparameter optimization and learning-to-learn

Evolution

Population Based Training of Neural Networks

利用meta-learning改进Bayesian optimization

Scalable Meta-Learning for Bayesian Optimization

Initializing Bayesian Hyperparameter Optimization via Meta-Learning

利用贝叶斯优化改进强化学习

Using Bayesian Optimization for Reinforcement Learning