Research map / Probabilistic and scientific modeling / Variational and generative sampling
Hindsight Network Credit Assignment: Efficient Credit Assignment in Networks of Discrete Stochastic Units
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use HNCA instead of REINFORCE | Bernoulli networks with softmax output · MNIST contextual bandit | test accuracy | Helped |
| Use HNCA instead of REINFORCE | Bernoulli networks with softmax output · MNIST contextual bandit | gradient variance | Helped |
| Subtract a moving-average reward baseline (ablation) | HNCA · MNIST contextual bandit | test accuracy | Mixed |
| Increase the network from one to more hidden layers (ablation) | HNCA and REINFORCE · MNIST contextual bandit | test accuracy | Hurt |
| Use f-HNCA instead of REINFORCE | Single-layer discrete VAE · MNIST | ELBO | Helped |
| Use f-HNCA with a layer-specific moving-average baseline (ablation) | Multi-layer discrete VAE · MNIST | ELBO | Helped |
| Use f-HNCA instead of the other tested estimators | Multi-layer discrete VAE · MNIST | ELBO | Hurt |
| Add a layer-specific moving-average baseline (ablation) | REINFORCE · Multi-layer discrete VAE | ELBO | Helped |
| Increase the number of stochastic layers (ablation) | f-HNCA · MNIST discrete VAE | ELBO | Mixed |
| Use HNCA instead of REINFORCE | A nonlinear network with a final Bernoulli hidden layer · MNIST contextual bandit | test accuracy | Helped |
| Exclude children with no downstream connection to a function component when computing credit (ablation) | f-HNCA with Baseline · MNIST discrete VAE | ELBO | Mixed |
| Exclude upstream function components from credit computation (ablation) | f-HNCA with Baseline · MNIST discrete VAE | ELBO | No clear effect |
Closest papers by meaning
- Structural Credit Assignment in Neural Networks using Reinforcement Learning (NeurIPS 2021 · similarity 0.91)
- Would I have gotten that reward? Long-term credit assignment by counterfactual contribution analysis (NeurIPS 2023 · similarity 0.89)
- Learning to solve the credit assignment problem (ICLR 2020 · similarity 0.88)
- Latent Reward: LLM-Empowered Credit Assignment in Episodic Reinforcement Learning (AAAI 2025 · similarity 0.86)
- STAS: Spatial-Temporal Return Decomposition for Solving Sparse Rewards Problems in Multi-agent Reinforcement Learning (AAAI 2024 · similarity 0.85)
- Retrospective In-Context Learning for Temporal Credit Assignment with Large Language Models (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.