2022 Developer Survey is open! Take survey.

Questions tagged [off-policy-methods]

For questions related to off-policy reinforcement learning algorithms, which estimate a policy (the target policy) while using another policy (the behavior policy), during the learning process, which ensures that all states are sufficiently explored. An example of an off-policy algorithm is Q-learning.

Filter by
Sorted by
Tagged with
1 vote
0 answers
42 views

Why do off-policy algorithms suffer from worse computational or time efficiency compared to on-policy algorithms?

When I run Soft-Actor-Critic (off-policy) in my Environment, the calculation of gradient updates takes almost twice the time compared to using PPO (on-policy). I also saw that ACER has a higher time ...
user avatar
  • 85
0 votes
0 answers
16 views

Can off-policy algorithms benefit from the parallelization?

On-policy algorithms, such as A2C, A3C and PPO, leverage massive parallelization to achieve state of the art results. However, I’ve never come across parallelization efforts when it comes to the off-...
user avatar
  • 167
1 vote
0 answers
47 views

Why does importance sampling ratio start and end one step later in off-policy SARSA given in Sutton-Barto's RL book?

In Sutton & Barto's RL book (page 149) they say: Sarsa update can be completely replaced by a simple off-policy form $Q_{t+n}(S_t,A_t)=Q_{t+n−1}(S_t,A_t) + \rho_{t+1:t+n} [G_{t:t+n} − Q_{t+n−1}(...
user avatar
1 vote
0 answers
37 views

Proper way to count environment steps / frames in distributed RL architecture for algorithms like CLEAR or LASER => modified impala with replay

In classical - on-policy - vtrace/Impala algorithm env_steps are incremented every training iteration like this : ...
user avatar
2 votes
1 answer
78 views

How does this TD(0) off-policy value update formula work?

The update formula for the TD(0) off-policy learning algorithm is (taken from these slides by D. Silver for lecture 5 of his course) $$ \underbrace{V(S_t)}_{\text{New value}} \leftarrow \underbrace{V(...
user avatar
  • 133
2 votes
1 answer
37 views

How to prove importance sampling ratio is uncorrelated with action-value (or state-value) estimate?

In Sutton & Barto (2nd edition), the following is mentioned on page 150 (p. 172 of the pdf), section 7.4: the importance sampling ratio has expected value one (Section 5.9) and is uncorrelated ...
user avatar
2 votes
0 answers
38 views

Doubt in Sutton & Barto's off-policy Monte Carlo control algorithm

The algorithm is described as below: My understanding: In the third last step, we act greedily w.r.t $Q$. Since we use importance sampling, this $Q \approx Q_\pi$. However, in the next step, whenever ...
user avatar
1 vote
1 answer
63 views

Which policy has to be followed by a player while construction of its own Q-table?

Consider the scenario, where there are two players. One of the players perform the action randomly, whereas I want second player as a Q-player. I mean, the player selects a best action from the Q-...
user avatar
  • 187
0 votes
1 answer
111 views

Which policy do I need to use in updating Q function?

Policy function can be of two types: deterministic policy and stochastic policy. Deterministic policy is of the form $\pi : S \rightarrow A$ Stochastic policy is defined using conditional probability ...
user avatar
  • 187
2 votes
1 answer
91 views

In off-policy MC control algorithm by Sutton & Barto, why do we perform a last update when sample action is inconsistent with target policy?

I have a question about the $W$ term in the off-policy MC control algorithm on Page 111 of Sutton & Barto. I have also included it in the figure below. My question: shouldn't the check $A_{t} = \...
user avatar
2 votes
2 answers
187 views

With Monte Carlo off-policy learning what do we correct by using importance sampling?

I do not understand the link of importance sampling to Monte Carlo off-policy learning. We estimate a value using sampling on whole episodes, and we take these values to construct the target policy. ...
user avatar
3 votes
1 answer
106 views

Why can we take the action $a$ from the next state $s'$ in the max part of the Q-learning update rule, if that action doesn't lead to any reward?

I'm using OpenAI's cartpole environment. First of all, is this environment not Markov? Knowing that, my main question concerns Q-learning and off-policy methods: For me, there is something weird in ...
user avatar
  • 145
2 votes
1 answer
171 views

Can I add expert data to the replay buffer used by the DDPG algorithm in order to make it converge faster?

I am working on a restricted reinforcement learning environment, i.e. the environment breaks very often (i.e.: the communication between the simulator and reinforcement learning agent breaks after ...
user avatar
2 votes
1 answer
277 views

Offline/Batch Reinforcement Learning: when to stop training and what agent to select

Context: My team and I are working on a RL problem for a specific application. We have data collected from user interactions (states, actions, rewards, etc.). It is too costly for us to emulate agents....
user avatar
  • 21
5 votes
1 answer
236 views

Why do we need importance sampling?

I was studying the off-policy policy improvement method. Then I encountered importance sampling. I completely understood the mathematics behind the calculation, but I am wondering what is the ...
user avatar
4 votes
1 answer
107 views

Why does off-policy learning outperform on-policy learning?

I am self-studying about Reinforcement Learning using different online resources. I now have a basic understanding of how RL works. I saw this in a book: Q-learning is an off-policy learner. An off-...
user avatar
3 votes
1 answer
101 views

When learning off-policy with multi-step returns, why do we use the current behaviour policy in importance sampling?

When learning off-policy with multi-step returns, we want to update the value of $Q(s_1, a_1)$ using rewards from the trajectory $\tau = (s_1, a_1, r_1, s_2, a_2, r_2, ..., s_n, a_n, r_n, s_n+1)$. We ...
user avatar
2 votes
0 answers
69 views

Is Deep SARSA learning a feasible approach?

I noticed that SARSA has been rarely used in the deep RL setting. Usually, the training for DQN is done off-policy. I think one of the major reasons for this is due to greater sample efficiency in ...
user avatar
  • 1,201
3 votes
0 answers
231 views

In Soft Actor Critic, why is the action sampled from current policy instead of replay buffer on value function update?

While reading the original paper of Soft Actor Critic, I came across on page number 5, under equation (5) and (6) $$ J_{V}(\psi)=\mathbb{E}_{\mathbf{s}_{t} \sim \mathcal{D}}\left[\frac{1}{2}\left(V_{\...
user avatar
0 votes
0 answers
48 views

Off-policy full-random training in easy-to-explore environment

Let say we are in an environment where a random agent can easily explore all the states of an environment (for example: tic-tac-toe). In those environments, using off-policy algorithm, is it a good ...
user avatar
  • 266
0 votes
1 answer
29 views

Learning only using off-policy samples

When training policies, is there a reason we need on-policy samples? For expensive simulations, it makes sense to try and reuse samples. Say we're interested in hyperparameter tuning. Can we collect a ...
user avatar
  • 101
1 vote
0 answers
42 views

How can I derive n-step off-policy temporal difference formula?

I was reading the book "Reinforcement Learning: An Introduction" by Sutton and Barto. In section 7.3, they write the formula for n-step off-policy TD as $$V(S_t) = V(S_{t-1}) + \alpha \rho_{...
user avatar
3 votes
1 answer
131 views

What is meant by "generate the data" in describing the difference between on-policy and off-policy?

From the book: Sutton, Richard S.,Barto, Andrew G.. Reinforcement Learning (Adaptive Computation and Machine Learning series) (p. 100). The MIT Press. Kindle Edition. " following is stated: "...
user avatar
  • 295
1 vote
0 answers
41 views

Should the importance sampling ratio be updated at the end of the for loop in the off-policy Monte Carlo control algorithm?

I'm studying RL with Sutton and Barto's book. I'd like to ask about the order of execution of a statement in the algorithm below. Here, $W$ (importance sampling ratio) is updated at the end of the <...
user avatar
  • 11
1 vote
0 answers
46 views

Does importance sampling for off-policy estimation also apply to the case of negative rewards?

Importance sampling is a common method for calculating off-policy estimates in RL. I have been reading through some of the original documentation (D.G. Horvitz and D.J. Thompson, Powell, M.J. and ...
user avatar
3 votes
0 answers
41 views

Why is it the case that off-policy evaluation using importance sampling suffers from high variance?

The average return for trajectories, $V^{\pi_e}$(s) is often computed via the importance sampling estimate $$V^{\pi_e}(s) = \frac{1}{n}\sum_{i=1}^n\prod_{t=0}^{H}\frac{\pi_e(a_t | s_t)}{\pi_b(a_t|s_t)}...
user avatar
  • 1,201
1 vote
0 answers
55 views

Is this figure a correct representation of off-policy actor-critic methods?

Does this figure correctly represent the overall general idea about actor-critic methods for on-policy (left) and off-policy (right) case? I am a bit confused about the off-policy case (right figure). ...
user avatar
1 vote
0 answers
84 views

Does the off-policy evaluation work for non-stationary policies?

As the title says, in reinforcement learning, does the off-policy evaluation work for non-stationary policies? For example, IS (importance sampling)-based estimators, such as weighted IS or doubly ...
user avatar
  • 207
0 votes
1 answer
110 views

How do I know that the DQN has learnt an appropriate Q function?

Is there any sanity check to know whether the Q functions learnt are appropriate in deep Q networks? I know that the Q values for end states should approximate the terminal reward. However, is it ...
user avatar
  • 1,201
2 votes
2 answers
506 views

How is per-decision importance sampling derived in Sutton & Barto's book?

In per-decison importance sampling given in Sutton & Barto's book: Eq 5.12 $\rho_{t:T-1}R_{t+k} = \frac{\pi(A_{t}|S_{t})}{b(A_{t}|S_{t})}\frac{\pi(A_{t+1}|S_{t+1})}{b(A_{t+1}|S_{t+1})}\frac{\pi(...
user avatar
2 votes
0 answers
54 views

Can weighted importance sampling be applied to off-policy evaluation for continuous state space MDPs?

Can weighted importance sampling (WIS) and importance sampling (IS) be applied to off-policy evaluation for continuous state spaces MDPs? Given that I have trajectories of $(s_t,a_t)$ pairs and the ...
user avatar
  • 1,201
2 votes
3 answers
237 views

What is the intuition behind importance sampling for off-policy value evaluation?

The technique for off-policy value evaluation comes from importance sampling, which states that $$E_{x \sim q}[f(x)] \approx \frac{1}{n}\sum_{i=1}^n f(x_i)\frac{q(x_i)}{p(x_i)},$$ where $x_i$ is ...
user avatar
  • 1,201
3 votes
1 answer
144 views

What is the difference between on-policy and off-policy for continuous environments?

I'm trying to understand RL applied to time series (so with infinite horizon) which have a continous state space and a discrete action space. First, some preliminary questions: in this case, what is ...
user avatar
  • 281
1 vote
1 answer
67 views

Is it possible to prove that the target policy is better than the behavioural policy based on learned Q values?

I have retrospective data for a sort of "behaviour policy" which I will use to train a deep q network to learn a target greedy policy. After learning the Q values for this target policy, can we make ...
user avatar
  • 1,201
1 vote
1 answer
182 views

Can we combine Off-Policy with On-Policy Algorithms?

On-Policy Algorithms like PPO directly maximize the performance objective or an approximation of it. They tend to be quite stable and reliable but are often sample inefficient. Off-Policy Algorithms ...
user avatar
2 votes
2 answers
566 views

Why is DDPG an off-policy RL algorithm?

In DDPG, if there are no $\epsilon$-greedy and no action noise, is DDPG an on-policy algorithm?
user avatar
2 votes
3 answers
90 views

How to estimate a behavior policy for off-policy learning based on data?

I have a dataset which includes states, actions, and reward. The dataset includes information on the transition, i.e., $p(r,s' \mid s,a)$. Is there a way to estimate a behavior policy from this ...
user avatar
1 vote
1 answer
444 views

What are the differences between 1-step SARSA and SARSA?

SARSA is on-policy, while n-step SARSA is off-policy. But when n = 1, is it like an off-policy version of SARSA? Any similarity and difference between 1-step SARSA and SARSA?
user avatar
1 vote
1 answer
66 views

Understanding the W term in off policy monte carlo learning

In Sutton and Barto's RL textbook they included the following pseudocode for off policy Monte Carlo learning. I am a little confused, however, because to me it looks like the W term will become ...
user avatar
6 votes
1 answer
2k views

Is Expected SARSA an off-policy or on-policy algorithm?

I understand that SARSA is an On-policy algorithm, and Q-learning an off-policy one. Sutton and Barto's textbook describes Expected Sarsa thusly: In these cliff walking results Expected Sarsa was ...
user avatar
  • 63
2 votes
1 answer
85 views

Could we update the policy network with previous trajectories using supervised learning?

I believe to understand the reason why on-policy methods cannot reuse trajectories collected from earlier policies: the trajectory distribution change with the policy and the policy gradient is ...
user avatar
2 votes
1 answer
43 views

Equivalence between expected parameter increments in "Off-Policy Temporal-Difference Learning with Function Approximation"

I am having a hard time understanding the proof of theorem 1 presented in the "Off-Policy Temporal-Difference Learning with Function Approximation" paper. Let $\Delta \theta$ and $\Delta \...
user avatar
2 votes
1 answer
924 views

Are model-free and off-policy algorithms the same?

In respect of RL, is model-free and off-policy the same thing, just different terminology? If not, what are the differences? I've read that the policy can be thought of as 'the brain', or decision ...
user avatar
  • 573
2 votes
0 answers
54 views

Can the importance sampling estimator have a non-stationary behaviour policy even if the target policy is stationary?

The inverse propensity score (IPS) estimator, which is used for off-policy evaluation in a contextual bandit problem, is well explained in the paper Doubly Robust Policy Evaluation and Optimization. ...
user avatar
  • 207
3 votes
1 answer
91 views

What is a non-starving policy in reinforcement learning?

In the paper, Eligibility Traces for off-Policy Policy Evaluation (2010), by Doina Precup et al., mentioned the term "non-starving" many times. The specific use of the term was like "non-starving ...
user avatar
  • 490
1 vote
1 answer
44 views

Do I need to store the policy for RL?

I am creating a zero-sum game with RL and wondered if I need to store the policy, or if there are other RL methods that produce similar results (consistently beating the human player) without the need ...
user avatar
  • 573
5 votes
1 answer
288 views

Understanding the n-step off-policy SARSA update

In Sutton & Barto's book (2nd ed) page 149, there is the equation 7.11 I am having a hard time understanding this equation. I would have thought that we should be moving $Q$ towards $G$, where $...
user avatar
2 votes
1 answer
84 views

Why is $M_t$ (the emphasis) helping in correcting for the state distribution in the Emphatic TD algorithm?

The book by Sutton and Barto discusses in section 11.8 that the convergence of off-policy TD function approximation can be improved by correcting for the distribution of states encountered. The ...
user avatar
  • 209
11 votes
1 answer
4k views

What is the relation between online (or offline) learning and on-policy (or off-policy) algorithms?

In the context of RL, there is the notion of on-policy and off-policy algorithms. I understand the difference between on-policy and off-policy algorithms. Moreover, in RL, there's also the notion of ...
user avatar
  • 33.1k
5 votes
1 answer
133 views

How do I compute the variance of the return of an evaluation policy using two behaviour policies?

Suppose there is an evaluation policy called $\pi_{e}$ and there are two behavior policies $\pi_{b1}$ and $\pi_{b2}$. I know that it is possible to estimate the return of policy $\pi_{e}$ through ...
user avatar
  • 451