Research map / Learning theory and algorithms / Prediction and decision losses
Towards a fuller understanding of neurons with Clustered Compositional Explanations
Biagio La Rosa, Leilani H. Gilpin, Roberto Capobianco
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add areas heuristic to CoEx (ablation) | CoEx · ResNet18, Ade20k | visited states: 39656 -> 23602 | Improved |
| add coordinates-free heuristic (CFH) to CoEx (ablation) | CoEx · ResNet18, Ade20k | visited states: 39656 -> 5990 | Improved |
| add label mask estimation (full MMESH) (ablation) | CoEx+CFH · ResNet18, Ade20k | visited states: 5990 -> 129 | Improved |
| cluster activations then apply CoEx instead of CoEx alone | CoEx · ResNet18, Ade20k | IoU: 0.08 -> 0.15 | Improved |
| cluster activations then apply CoEx instead of CoEx alone | CoEx · ResNet18, Ade20k | ExplCov: 0.14 -> 0.60 | Improved |
| cluster activations then apply CoEx instead of CoEx alone | CoEx · ResNet18, Ade20k | DetAcc: 0.18 -> 0.20 | Improved |
| cluster activations then apply CoEx instead of CoEx alone | CoEx · ResNet18, Ade20k | LabMask: 0.57 -> 0.61 | Mixed |
| use highest-activation Cluster 5 versus CoEx range | CoEx · ResNet18, Ade20k | IoU: 0.08 -> 0.09 | No clear effect |
| use highest-activation Cluster 5 versus CoEx range | CoEx · ResNet18, Ade20k | ActCov: 0.15 -> 0.22 | Improved |
| increase number of clusters from 5 to 10 (ablation) | Clustered CE · ResNet18, Ade20k | IoU: 0.15 -> 0.09 | Hurt |
| increase number of clusters from 5 to 15 (ablation) | Clustered CE · ResNet18, Ade20k | IoU: 0.15 -> 0.07 | Hurt |
| increase number of clusters from 5 to 15 (ablation) | Clustered CE · ResNet18, Ade20k | DetAcc: 0.21 -> 0.09 | Hurt |
Closest papers by meaning
- Evaluating Neuron Explanations: A Unified Framework with Sanity Checks (ICML 2025 · similarity 0.87)
- Linear Explanations for Individual Neurons (ICML 2024 · similarity 0.84)
- Learning to cluster neuronal function (NeurIPS 2025 · similarity 0.84)
- Spectral Analysis of Representational Similarity with Limited Neurons (NeurIPS 2025 · similarity 0.83)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.83)
- When does compositional structure yield compositional generalization? A kernel theory. (ICLR 2025 · similarity 0.83)
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.
