In Reinforcement Learning, epsilon-greedy policies are the most used exploration policies, but in case there is a big state space (let's consider 100,000 states and 2,000 actions) with impossible actions (if this occurs the agent goes to a dummy state and gets a bad reward), wouldn't it be better to use soft-max policies instead?

In Reinforcement Learning, epsilon-greedy policies are the most used exploration policies, but in case there is a big state space with impossible actions, wouldn't it be better to use soft-max policies instead?

In Reinforcement Learning, epsilon-greedy policies are the most used exploration policies, but in case there is a big state space (let's consider 100,000 states and 2,000 actions) with impossible actions (if this occurs the agent goes to a dummy state and gets a bad reward), wouldn't be better to use soft-max policies?

In Reinforcement Learning, epsilon-greedy policies are the most used exploration policies, but in case there is a big state space (let's consider 100,000 states and 2,000 actions) with impossible actions (if this occurs the agent goes to a dummy state and gets a bad reward), wouldn't it be better to use soft-max policies instead?

In Reinforcement Learning epsilon-greedy policies are the most used exploration policies, but in case there is a big state space (let's consider 100,000 states and 2,000 actions) with impossible actions (if this occurs the agent goes to a dummy state and gets a bad reward), wouldn't be better to use soft-max policies?

In Reinforcement Learning, epsilon-greedy policies are the most used exploration policies, but in case there is a big state space (let's consider 100,000 states and 2,000 actions) with impossible actions (if this occurs the agent goes to a dummy state and gets a bad reward), wouldn't be better to use soft-max policies?