Research map / Vision-language and multimodal / Human and embodied modeling
Feature segregation by signed weights in artificial vision systems and biological models
Giordano Ramos-Traslosheros, Carlos R Ponce
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 |
|---|---|---|---|
| remove positive weights to class units (ablate positive inputs) (ablation) | ImageNet CNN output units · Background challenge | background reliance R_bg: 0.157 -> 0.488 | Hurt |
| remove negative weights to class units (ablation) | ImageNet CNN output units · Background challenge | background reliance R_bg: 0.157 -> 0.078 | Improved |
| ablate positive weights (ablation) | ImageNet CNN output units · Background challenge | foreground reliance R_fg: 0.903 -> 0.839 | No clear effect |
| ablate negative weights (ablation) | ImageNet CNN output units · Background challenge | foreground reliance R_fg: 0.903 -> 0.711 | Hurt |
| use ReLU instead of Tanh activations (ablation) | ResNet18 trained 16 epochs · ImageNet-1k val | top-5 accuracy: 0.797 -> 0.870 | Improved |
| use ReLU instead of Tanh activations (ablation) | ResNet18 · feature segregation by sign | representational change under ablation: ReLU model showed clear segregation; Tanh similar changes across ablation types | Improved |
| add adversarial robust training (larger L∞) (ablation) | ResNet50 · ablation similarity change | Spearman rho of change vs robustness: -0.47 (6e-5) positive, -0.57 (5e-7) negative at alpha 1.0 | Improved |
| train with supervision and more depth (training vs random init) | ImageNet CNNs · output layer | Dale index: 0.6004 vs 0.5125 for resnet50 | Improved |
| add batchnorm to VGG | VGG16 · ImageNet-1k val | Dale index: 0.5606 -> 0.5761 | Improved |
| use unsupervised SimSiam pretraining with linear head | ResNet50 · feature segregation | ablation effect: object features still in positive weights; disappeared at lower ablation strengths | No clear effect |
| constrain neuron models to positive input weights only (ablation) | PLS neuron models of macaque V1/V4/IT · neural response prediction | train and test accuracy | Hurt |
| clear the background around neuron's preferred feature | macaque ventral stream neurons · in vivo | firing response | Improved |
Closest papers by meaning
- Anatomically inspired digital twins capture hierarchical object representations in visual cortex (NeurIPS 2025 · similarity 0.88)
- Generalized Shape Metrics on Neural Representations (NeurIPS 2021 · similarity 0.87)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · similarity 0.86)
- Parallel Backpropagation for Shared-Feature Visualization (NeurIPS 2024 · similarity 0.86)
- Bridging Critical Gaps in Convergent Learning: How Representational Alignment Evolves Across Layers, Training, and Distribution Shifts (NeurIPS 2025 · similarity 0.85)
- Scaling Laws for Task-Optimized Models of the Primate Visual Ventral Stream (ICML 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.
