Distilling Reinforcement Learning

The paper introduces a hierarchical reinforcement learning framework called Feudal Networks (FuN) that enables agents to learn a hierarchy of temporal abstractions. The proposed model consists of a Manager and a Worker, with the Manager learning high-level policies and the Worker learning low-level policies. The approach is demonstrated to achieve state-of-the-art performance on a range of challenging tasks in the Atari domain.

In machine learning, boosting is an ensemble meta-algorithm for primarily reducing bias, and also variance in supervised learning, and a family of machine learning algorithms that convert weak learners to strong ones. Wikipedia

Naive Bayes classifiers are highly scalable, requiring a number of parameters linear in the number of variables (features/predcitors) in a learning problem. Maxumum-likelihood training can be done by evaluting a closed-form exporession, which takes linear time, rather tahn by expensive iterative approximation as used for many other typs of classifier. Wikipedia

If you ask yourself what’s the most important thing in machine learning, what’s your answer? All data scientist would have different answers.

To protect our system and computer we should make sure that data which we download is clean. Everytime we bring data to our system or user upload data such as file attachments, we must make sure that data is free from viruses and trojans.

SFTP (SSH File Transfer Protocol) is a secure file transfer protocol. It runs over the SSH protocol. It supports the full security and authentication functionality of SSH.

Ananyzing the ouput of a simulation model is important. How can we be sure that our output is proper and will not hurt an experiment result using those outputs.

How can we tell our random variables are well made? In simulation terminology, we have something called input analysis.