Skip to content

Update README.md #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
gunasekhar159 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Update README.md #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
67 changes: 59 additions & 8 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,22 +2,73 @@


## AIM
Write the experiment AIM.
To develop a Python program to find the optimal policy for the given RL environment using Q-Learning and compare the state values with the Monte Carlo method.

## PROBLEM STATEMENT
Explain the problem statement.
Develop a Python program to derive the optimal policy using Q-Learning and compare state values with Monte Carlo method.

## Q LEARNING ALGORITHM
Include the steps involved in the Q Learning algorithm
Step 1:
Initialize Q-table and hyperparameters.

Step 2:
Choose an action using the epsilon-greedy policy and execute the action, observe the next state, reward, and update Q-values and repeat until episode ends.

Step 3:
After training, derive the optimal policy from the Q-table.

Step 4:
Implement the Monte Carlo method to estimate state values.

Step 5:
Compare Q-Learning policy and state values with Monte Carlo results for the given RL environment.
## Q LEARNING FUNCTION
Include the Q Learning function
Developed by:M.Gunasekhar
Ref no :212221240014

def q_learning(env,
gamma=1.0,
init_alpha=0.5,
min_alpha=0.01,
alpha_decay_ratio=0.5,
init_epsilon=1.0,
min_epsilon=0.1,
epsilon_decay_ratio=0.9,
n_episodes=3000):
nS, nA = env.observation_space.n, env.action_space.n
pi_track = []
Q = np.zeros((nS, nA), dtype=np.float64)
Q_track = np.zeros((n_episodes, nS, nA), dtype=np.float64)
select_action=lambda state,Q,epsilon: np.argmax(Q[state]) if np.random.random()>epsilon else np.random.randint(len(Q[state]))
alphas=decay_schedule(
init_alpha,min_alpha,
alpha_decay_ratio,
n_episodes)
epsilons=decay_schedule(
init_epsilon,min_epsilon,
epsilon_decay_ratio,
n_episodes)
for e in tqdm(range(n_episodes),leave=False):
state,done=env.reset(),False
action=select_action(state,Q,epsilons[e])
while not done:
action=select_action(state,Q,epsilons[e])
next_state,reward,done,_=env.step(action)
td_target=reward+gamma*Q[next_state].max()*(not done)
td_error=td_target-Q[state][action]
Q[state][action]=Q[state][action]+alphas[e]*td_error
state=next_state
Q_track[e]=Q
pi_track.append(np.argmax(Q,axis=1))
V=np.max(Q,axis=1)
pi=lambda s:{s:a for s,a in enumerate(np.argmax(Q,axis=1))}[s]
return Q, V, pi, Q_track, pi_track
## OUTPUT:
Mention the optimal policy, optimal value function , success rate for the optimal policy.

Include plot comparing the state value functions of Monte Carlo method and Qlearning.
<img width="636" alt="image" src="https://github.com/obedotto/q-learning/assets/95043391/7e4ce389-307d-40b8-b2d2-dcecd5d727aa">
<img width="479" alt="image" src="https://github.com/obedotto/q-learning/assets/95043391/8c55f097-cdc5-4eea-a22f-34137259ecdd">
<img width="664" alt="image" src="https://github.com/obedotto/q-learning/assets/95043391/3a1b1b17-fb47-4bc1-831b-410d698a0d03">
<img width="671" alt="image" src="https://github.com/obedotto/q-learning/assets/95043391/330d30ea-53e4-4674-b3ef-72a04d7f3626">

## RESULT:
Therefore a python program has been successfully developed to find the optimal policy for the given RL environment using Q-Learning and compared the state values with the Monte Carlo method.

Write your result here