Why does policy gradiet method has high variance?
策略梯度方法 策略梯度方法中，目標函數是使得整個episode得到的reward的均值最大： maximizeθEπθ[∑t=0T−1γtrt] 由於： ∇θE[f(x)]=∇θ∫pθ(x)f(x)dx=∫pθ(x)pθ(x)∇θpθ(x)f(x)dx=∫pθ(x)∇θlogpθ(x)f(x)dx=E[f(x)∇θlogpθ(x)] 以及： ∇θlogpθ(τ)=∇log(μ(s0)∏t=0T−1
相關文章
- 1. Why does deep learning work?
- 2. why request method is OPTIONS
- 3. Why does Double.NaN==Double.NaN return false?
- 4. A Policy Update Strategy in Model-free Policy Search: Policy Gradient
- 5. ModelMapper報錯Ensure that method has zero parameters and does not return void.
- 6. Policy Gradient Algorithms
- 7. （轉）RL — Policy Gradient Explained
- 8. Why UI correction note always has a big static size
- 9. Where does the error come from?----Bias and Variance
- 10. Privacy Policy
- 更多相關文章...
- • PHP range() 函數 - PHP參考手冊
- • WebSecurity - UserExists() - ASP.NET 教程
- • Flink 數據傳輸及反壓詳解
- • Spring Cloud 微服務實戰(三) - 服務註冊與發現
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. Why does deep learning work?
- 2. why request method is OPTIONS
- 3. Why does Double.NaN==Double.NaN return false?
- 4. A Policy Update Strategy in Model-free Policy Search: Policy Gradient
- 5. ModelMapper報錯Ensure that method has zero parameters and does not return void.
- 6. Policy Gradient Algorithms
- 7. （轉）RL — Policy Gradient Explained
- 8. Why UI correction note always has a big static size
- 9. Where does the error come from?----Bias and Variance
- 10. Privacy Policy