Here is a useful website for starters to understand reinforcement learning: https://missinglink.ai/guides/deep-learning-healthcare/tensorflow-reinforcement-learning-introduction-and-hands-on-tutorial/

Useful Website for Reinforcement learning

In reinforcement learning, the environment encompasses everything external to the agent with which it interacts. It processes the agent's action in its current state and responds by providing a reward and transitioning the agent to a new state. For example, in a physical system, the environment could be the laws of physics. From the agent's perspective, the environment often functions as a black box.

Environment in Reinforcement Learning

In reinforcement learning, a state (s) represents the current situation or configuration of the environment. It is a snapshot of all relevant information at a specific moment that the agent uses to make a decision. This information can include the agent's position, the status of other entities, and any other data that defines the current circumstances.

State in Reinforcement Learning

In reinforcement learning, the agent is the component that functions as the learner or decision-maker. It interacts with an environment by perceiving its state, performing actions, and learning from the resulting feedback. For instance, an agent could be a robot navigating a path or a trading algorithm making financial decisions. In the context of Large Language Models (LLMs), the LLM itself often serves as the agent.

Agent in Reinforcement Learning

An action is the set of all possible moves the agent can make. Agents choose from a list of possible actions. In video games, for example, the list might include: running right or left, jumping high or low, crouching or standing still.

Action in Reinforcement Learning

In reinforcement learning, a reward is a signal sent from the environment to the agent that provides feedback on an action's success. This feedback, which can be positive or negative, guides the agent's learning process by indicating the desirability of the actions taken. The agent's primary objective is to modify its policy over time to maximize the cumulative reward it receives.

Reward in Reinforcement Learning

Here is a free pdf version of a book about Reinforcement Learning:
http://incompleteideas.net/book/RLbook2020.pdf

Useful Book for Reinforcement Learning

https://cims.nyu.edu/~donev/Teaching/WrittenOral/Projects/XintianHan-WrittenAndOral.pdf
https://www.davidsilver.uk/teaching/     
(David Silver's tutorial of reinforcement learning)
http://rail.eecs.berkeley.edu/deeprlcourse/
https://katefvision.github.io/
http://incompleteideas.net/rlai.cs.ualberta.ca/RLAI/RLAIcourse/RLAIcourse2006.html
http://web.stanford.edu/class/cs234/schedule.html

Useful Tutorials about Math behind Reinforcement Learning

Reinforcement learning is a general approach to solve reward-based problem. The building of it is based on mathematical foundations. We can also view it as a sequential decision making and control problem mathematically. For example, when driving a car, we need to make decisions about turning left/right or going straight after the previous decisions. And the key ingredient in reinforcement learning is Markov Decision Process.

Math Behind Reinforcement Learning

It is very challenging in the reinforcement learning to find a perfect balance between using previously used effective action(**exploitation**) and exploring new ones (**exploration**).

In order to solve this problem, simple $\epsilon$-Greedy algorithm can be used, in which we with some small probability $\epsilon$ we would select action randomly (**exploration**) and with $1 - \epsilon$ probability use previously selected effective action(**exploitation**).  

Exploration/Exploitation trade-off

Classification of Reinforcement Learning Methods

On-policy methods attempt to evaluate or improve the policy that is used to make decisions. In contrast, off-policy methods evaluate or improve a policy different from that used to generate the data. Classicla on-policy methods include Sarsa Algorithm. As a comparison, Q-learning method is a classical off-policy method.

On-policy vs Off-policy

Compared to value based methods, actor-critic methods focus on modeling the probability distribution of policies. For value-based methods, the logic is more like a greedy algorith, we want to maximize the value function for every step. As a result, we need to search through the action space to get the best action. But in actor-critic method, they are like using actor to perform actions based on the probability distribution of policies and using critics to judge the performance and adjust the probability distribution. They don't directly always choose the best action. 

Actor-Critic Methods

van Hasselt, H., Guez, A., & Silver, D. (2016). Deep Reinforcement Learning with Double Q-Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 30(1). Retrieved from https://paperswithcode.com/paper/deep-reinforcement-learning-with-double-q

Deep Reinforcement Learning with Double Q-learning

One of the most popular reinforcement learning algorithms, known to sometimes learn unrealistically high action values since it includes a maximization step that tends to prefer overestimated values.

Q-learning

Borges, Alexandre and Arlindo L. Oliveira. “Combining Off and On-Policy Training in Model-Based Reinforcement Learning.” ArXiv abs/2102.12194 (2021): n. pag.
https://arxiv.org/pdf/2102.12194.pdf

Combining Off and On-Policy Training in Model-Based
Reinforcement Learning

MuZero is a model-based reinforcement learning method which contains a representation function, a dynamics function, and a prediction function.The representation function encodes an observation sequence into a hidden state. The dynamics function evaluates the current hidden state and a potential action to determine the potential next hidden state and a reward value. Lastly, the prediction function is the policy generation function. They use similar parameters and are optimized according to a combined loss function.

A Monte Carlo Tree Search (MCTS) is often used when training to evaluate action trees from the current/ root state.

MuZero

In a reinforcement learning framework, the process of training a Large Language Model (LLM) iteratively evaluates and improves the model's policy. At each step $$t$$, the current state $$s_t$$ is defined by the initial input prompt $$\mathbf{x}$$ and the tokens generated so far $$\mathbf{y}_{<t}$$. The LLM acts as the policy, denoted by the predicted distribution $$\Pr(y_t|\mathbf{x},\mathbf{y}_{<t})$$, to choose an action $$a_t$$, which is the next token $$y_t$$. After $$y_t$$ is predicted, a reward model evaluates the sequence $$(\mathbf{x},\mathbf{y}_{<t}, y_t)$$ to determine how well it aligns with the desired textual outcome. This evaluation produces reward scores that are then used to compute the value functions $$V(s_t)$$ and $$Q(s_t,a_t)$$. Finally, these value functions provide the necessary feedback to guide the subsequent training and refinement of the LLM's policy.

Reinforcement Learning Process for LLMs

A robot is being trained to find the exit in a maze. At any given position, the robot can choose to move forward, turn left, or turn right. The system keeps track of the robot's current coordinates within the maze. The robot's objective is to reach the exit as efficiently as possible. It receives a large positive signal for reaching the exit, a large negative signal for hitting a wall, and a small negative signal for each step it takes. Based on this scenario, identify the five fundamental components of this learning system and briefly explain the specific role of each.

Analyzing a Learning System

A robot is being trained to navigate a maze to find a piece of cheese. Analyze this scenario by matching each element of the training process to its corresponding fundamental concept.

The core of reinforcement learning is the interaction between an agent and a dynamic environment, which is modeled as a sequential process. At every time step, the agent assesses the environment's current state and uses its policy to select an action. After executing the action, the environment provides feedback consisting of a reward and a new state. This cycle of observing, acting, and receiving feedback continues until the agent accomplishes its objective.

Agent-Environment Interaction Loop in Reinforcement Learning

A cat is learning to use a new automated feeder that dispenses food when a lever is pressed. Initially, the cat paws at the lever randomly. After several attempts, it presses the lever and food is dispensed. The cat begins to press the lever more frequently. Which of the following statements best analyzes the relationship between the core components in this learning scenario?

The fundamental concepts that form the basis of reinforcement learning are:

 - Agent
 - Environment
 - Action
 - State
 - Reward

The agent learns through repeated interaction with the environment. To be successful, the agent needs to:
 - Learn the interaction between states, actions, and subsequent rewards.
 - Determine which action will provide the optimal outcome.

Google

University of Michigan - Ann Arbor

In reinforcement learning, the machine goes through trial and error processes where it is rewarded/penalized for the actions it performs. The machine's goal is to maximize the total reward by leveraging previous attempts to make the next decision. The model starts from completely random trials and eventually leads to something with very sophisticated tactics and skills. Compared to supervised/unsupervised learning, reinforcement learning is much more advanced. A reinforcement model will continuously learn, unlike supervised/unsupervised models which all have an endpoint after the training and test data phases.

Reinforcement Learning

Reference of Foundations of Large Language Models Course

A good way to understand reinforcement learning would be to think about how autonomous vehicles work. Instead of programming all of the rules it should follow (such as road signs, pedestrians, right of way, etc), it is much more easier and efficient to implement a reward/penalty system where the car learns all of the rules that it is supposed to abide by. Furthermore, it would be unreasonable to program individual rules because accidents can happen in unpredictable ways and rules/regulations are always changing and varies in different parts of the world. So, it is much better for autonomous cars to be always learning better ways to make decisions.

Reinforcement Learning Example - Autonomous Vehicles

Video games often involve completing levels, with increasing difficulty as you go through the game. In each level, you go in completely blind and make decisions that will either lead to you passing that level or failing. If you fail, you will now understand more about that particular level that will influence your decisions when you retry. In other words, you will understand what not to do, and try another approach. If you pass, you will understand what to do in future levels. This process is repeated until success. This is very similar to how reinforcement learning works. Rewards are given for successes, indicating that the decision(s) made are correct and should be learned for future use. Penalties are given for a failure, indicating that the decision(s) made should not be retried and instead a new approach should be used.

Reinforcement Learning Analogy - Video Games

Theobald, O. (2017). Machine learning for absolute beginners: a plain English introduction (2nd ed.). Scatterplot Press.

Online pdf: http://bmansoori.ir/book/Machine%20Learning%20For%20Absolute%20Beginners.pdf

Machine Learning for Absolute Beginners

Deep learning analyses a training set, identifies complex patterns and applies them to new data. 

Reinforcement learning works sequentially in an unknown environment━taking an action, evaluating the rewards, and adjusting the following actions accordingly.

Deep learning and reinforcement learning complement each other:

 - Reinforcement learning algorithms manage the sequential process of taking an action, evaluating the result, and selecting the next best action. However, they need a good mechanism to select the best action based on previous interactions.
 - Deep learning can be that mechanism━it is the most powerful method available today to learn the best outcome based on previous data.

 

Deep Learning vs. Reinforcement Learning

Fundamental Concepts for Reinforcement Learning

(Encourage to continuously enrich and edit this node for sharing papers and their proposed useful ideas.)
1.Learning Action Representations for Reinforcement Learning Yash Chandak, Georgios Theocharous, James E. Kostas, Scott M. Jordan, Philip S. Thomas 

Reinforcement Learning Refrence and Cutting-edge Ideas

A team is developing a program to play a complex board game against human opponents. The program has no pre-existing data of past games to learn from. Instead, it is designed to learn by playing against itself repeatedly. After each game, the program receives a positive signal if it wins and a negative signal if it loses. Over time, it is expected to discover winning strategies on its own. Which of the following statements best analyzes why this learning approach is suitable for this task?

This scenario describes a classic reinforcement learning problem. Identify the Agent, Environment, State, Action, and Reward within this context. Then, explain how the reward structure incentivizes the agent to achieve the desired outcome of finding the exit quickly.

Robot Maze Navigation Strategy

A company is developing a system to recommend personalized news articles. The system's goal is to continuously adapt to a user's evolving interests. Two strategies are proposed:

**Strategy A:** Collect a massive, one-time dataset of user profiles and the articles they clicked on in the past. Train a model on this static dataset to predict which articles a new user is likely to click.

**Strategy B:** Design a system that learns through direct interaction. For each article it recommends, it receives a positive signal if the user clicks it and a negative signal if they ignore it. The system's objective is to maximize the positive signals it receives over time.

Evaluate these two strategies. Which one is better suited for creating a system that adapts to a user's *evolving* interests? Justify your evaluation by comparing the core mechanics of each strategy and their implications for long-term personalization.

Learn Before

Related

Learn After