AI/ML, Robotics

What do RL, Deep RL and IRL mean?

RL(Reinforcement learning) – An agent takes actions, and learns by maximizing the reward.

Deep RL – The model for an agent to take action is formed by a deep neural network. This enables the agent to have more complex behaviors.

IRL (Inverse Reinforcement learning) – An agent learns the reward function from the experiences.

Definition:

An easy-to-understand example:

AI/ML, Robotics

What does end-to-end learning mean?

End-to-end learning means you can learn the model like a black box. That is, you only need to take care of the input and the output and ignore the processing/abstraction in the middle.

This concept is illustrated in the picture below. Take computer vision pipeline as an example, you don’t need to take care of the mid-level features and classifiers if you use end-to-end learning.

end-tp-end

The picture above comes from Sergey Levine.

AI/ML

XGBoost演算法理解

這次因緣際會地學習了一點 XGBoost 的演算法概念,在看了許多部落格文章和網路上的資源後,覺得還是原作者講解的比較清楚且完整,覺得可以在這邊點出來,幫助更多人學懂這個好用的算法。

建議可以先專注地聽到 37 分鐘的地方,你大概就可以了解整個 XGBoost 的演算法精神了!很值得的半小時!