Add dueling in rainbow iqn#137
Closed
Curt-Park wants to merge 1 commit into
Closed
Conversation
Collaborator
Author
Curt-Park
added a commit
to Curt-Park/rainbow-is-all-you-need
that referenced
this pull request
Mar 18, 2026
Dueling architecture empirically degrades IQN performance (see medipixel/rl_algorithms#137). Replace the separate advantage/value streams with a single NoisyLinear output head. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Curt-Park
added a commit
to Curt-Park/rainbow-is-all-you-need
that referenced
this pull request
Mar 18, 2026
* Add Rainbow IQN tutorial (09) replacing C51 with Implicit Quantile Networks Introduces tutorial 09 which swaps the Categorical DQN (C51) distributional component in Rainbow with IQN. The network learns a continuous quantile function via cosine basis embeddings and Hadamard product, trained with quantile Huber loss — eliminating the need to specify v_min, v_max, or atom_size. All other Rainbow components (Double DQN, PER, Dueling, NoisyNet, N-step) remain unchanged. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Remove dueling network from Rainbow IQN Dueling architecture empirically degrades IQN performance (see medipixel/rl_algorithms#137). Replace the separate advantage/value streams with a single NoisyLinear output head. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Fix quantile Huber loss reduction to mean over both tau dimensions The IQN loss averages over both tau and tau_prime, but the previous code used sum over tau — inflating the loss ~32x and causing unstable training. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Restore dueling network in Rainbow IQN The earlier instability was caused by the loss reduction bug (sum instead of mean over tau), not the dueling architecture itself. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add Rainbow IQN tutorial to README Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Test and update session --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The test result:
https://app.wandb.ai/curt-park/dqn/runs/tftmxcjm?workspace=user-curt-park