Research map / Reinforcement learning / Policy gradient and actor-critic
Distinguishing Learning Rules with Brain Machine Interfaces
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 |
|---|---|---|---|
| Compare observed activity-flow changes with predictions based on the candidate learning rule | BMI cursor-control RNNs trained with biased supervised learning or node-perturbation reinforcement learning · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Helped |
| Increase alignment of the learned credit-assignment mapping with the decoder during training (ablation) | Supervised-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Hurt |
| Add driving feedback aligned with the supervised learner's credit-assignment mapping (ablation) | Supervised-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | No clear effect |
| Add strong driving feedback aligned with the credit-assignment mapping (ablation) | Reinforcement-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Hurt |
| Estimate the credit-assignment mapping from activity using PCA and linear regression instead of supplying the true mapping (ablation) | FFCC analysis of supervised-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | No clear effect |
| Use non-isotropic noise with very low dimensionality (ablation) | Supervised-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Hurt |
| Use non-isotropic noise with very low dimensionality (ablation) | Reinforcement-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | No clear effect |
| Increase the dimensionality of non-isotropic noise (ablation) | Supervised- and reinforcement-learning RNNs · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Helped |
| Assume isotropic noise when predicting activity changes despite non-isotropic actual noise (ablation) | FFCC analysis of reinforcement-learning RNN · Simulated BMI cursor-control task | FFCC-based learning-rule identification | No clear effect |
| Use decoder-aligned credit assignment instead of a sufficiently distinct, biased mapping (ablation) | FFCC analysis of supervised- and reinforcement-learning RNNs · Simulated BMI cursor-control task | FFCC-based learning-rule identification | Hurt |
Closest papers by meaning
- Flexible inference for animal learning rules using neural networks (NeurIPS 2025 · similarity 0.88)
- Can Biologically Plausible Temporal Credit Assignment Rules Match BPTT for Neural Similarity? E-prop as an Example (ICML 2025 · similarity 0.88)
- Dynamics of Supervised and Reinforcement Learning in the Non-Linear Perceptron (NeurIPS 2024 · similarity 0.87)
- Feedback control guides credit assignment in recurrent neural networks (NeurIPS 2024 · similarity 0.87)
- Model Based Inference of Synaptic Plasticity Rules (NeurIPS 2024 · similarity 0.85)
- Flexible task abstractions emerge in linear networks with fast and bounded units (NeurIPS 2024 · 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.