Problems

1 / 1

ε-Greedy Action Selection

Reinforcement Learning

Medium

Implement ε-greedy action selection for exploration in reinforcement learning.

Mathematical Definition

ε-Greedy Policy:

a = \begin{cases} \arg\max_{a} Q(s, a), & \text{with probability } 1 - \varepsilon, \\[6pt] \text{random action}, & \text{with probability } \varepsilon. \end{cases}

where ε∈[0,1] controls the exploration rate.

Function Arguments

q_values: 1D array, shape (n_actions,) - Q(s, ·) for current state
epsilon: float in [0,1] - exploration probability
rng: optional np.random.Generator - for deterministic testing

Loading visualization...

Examples

Input: q_values=[1,2,0.5], ε=0

Output: action = 1

Input: q_values=[1,2,0.5], ε=1

Output: action ∈ {0,1,2} with equal probability

Hint 1

Generate a random number between 0 and 1. If it's less than ε, choose random action, else choose greedy.

Hint 2

Use np.argmax() for greedy action and rng.integers() or np.random.randint() for random action.

Requirements

If ε=0 always return greedy action (argmax)
If ε=1 always return random action
Use rng if provided, else use np.random
Return Python int action index
Do not modify q_values

Constraints

1 ≤ len(q_values) ≤ 10,000
NumPy only
Time limit: 200ms

Try Similar Problems

Q Learning Update Sarsa Update Replay Buffer Sample Mc Policy Evaluation Td Value Update

Prev Next