Research map / Detection and segmentation / Open-set and few-shot detection
On Numerosity of Deep Neural Networks
Xi Zhang, Xiaolin Wu
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 |
|---|---|---|---|
| Increase the ANOVA test sample size from the original small sample | Object recognition DCNN · Dot displays | numerosity-selective neurons: 9.6% -> about 2% (20 images per combination); zero (80 images per combination) | Hurt |
| Use the original small ANOVA sample on randomly initialized weights | Untrained DCNN · Dot displays | numerosity-selective neurons: 12.8% (untrained) vs 9.6% (object recognition network) | Mixed |
| Train the DCNN specifically for numerosity classification | Nu-Net · Numerosity test, same distribution | accuracy: 95.6% | Improved |
| Train the DCNN specifically for numerosity classification | Nu-Net · Numerosity test, same distribution | numerosity-selective neurons: 68.4% | Improved |
| Increase object-size variation in test images by 50% relative to training | Nu-Net · Numerosity test | accuracy: 95.6% -> 64.2% | Hurt |
| Test with object sizes outside the training range | Nu-Net · Numerosity test, distribution shift | accuracy | Mixed |
| Train the Nu-Net with supervised numerosity labels | Nu-Net · Distribution-shifted numerosity, numbers 1 to 4 | accuracy | Improved |
Closest papers by meaning
- Cognitive Deficit of Deep Learning in Numerosity (AAAI 2019 · similarity 0.87)
- Machine Number Sense: A Dataset of Visual Arithmetic Problems for Abstract and Relational Reasoning (AAAI 2020 · similarity 0.83)
- Unveiling the Visual Counting Bottleneck in Vision-Language Models (ICML 2026 · similarity 0.81)
- Position: Deep Learning is Not So Mysterious or Different (ICML 2025 · similarity 0.81)
- Number Cookbook: Number Understanding of Language Models and How to Improve It (ICLR 2025 · similarity 0.81)
- Alignment of CNN and Human Judgments of Geometric and Topological Concepts (AAAI 2025 · similarity 0.81)
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.
