Reinforcement Learning

Question 1

Base components of every RL framework

Answer 1

Action, environment(state), reward

Question 2

What is a discount rate?

Answer 2

The discount factor essentially determines how much the reinforcement learning agents cares about rewards in the distant future relative to those in the immediate future.

Question 3

What is a policy?

Answer 3

A factory mapping states to actions, Often denoted using PI

Question 4

What is a greedy policy?

Answer 4

A policy where the agent always chooses the best expected return

Question 5

What is a discrete space?

Answer 5

Discrete spaces has finite states and finite actions

Question 6

What is a continuous space?

Answer 6

Continuous spaces can have a wide range of numbers and most physical space actions are continuous by nature

Question 7

When do you want to use MDPs like Monte carlo/TD-learning?

Answer 7

For finite spaces where actions are limited

Question 8

When do you want to use Deep reinforcement learning?

Answer 8

For continuous space tasks