Research map / Architectures and efficiency / RNNs and normalization
Polarity Is All You Need to Learn and Transfer Faster
Qingyang Wang, Michael Alan Powell, Eric W Bridgeford, Ali Geisa, Joshua T Vogelstein
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 |
|---|---|---|---|
| fix weight polarities a priori to an adequate (sufficient) template | Frozen-Net on XOR-5D · XOR-5D, 60-72 training samples | training time to reach same accuracy: takes 58% and 48% of Fluid-Net time | Improved |
| fix randomly chosen weight polarities (Bernoulli 0.5) | Frozen-Net RAND-Polarity on XOR-5D · XOR-5D | training time and data to reach accuracy | No clear effect |
| transfer ImageNet polarity pattern and fix it (Frozen-Net IN-Polarity) | AlexNet on Fashion-MNIST and CIFAR-10 · Fashion-MNIST and CIFAR-10, 100 training images | validation error at convergence: 7% (Fashion-MNIST), 9.4% (CIFAR-10) less error | Improved |
| transfer polarity pattern without freezing (Fluid IN-Polarity) vs Fluid-Net (ablation) | AlexNet · Fashion-MNIST and CIFAR-10 | validation error at convergence: 6% (Fashion-MNIST), 8.4% (CIFAR-10) gain | Improved |
| add polarity freezing on top of IN-Polarity transfer (ablation) | AlexNet · CIFAR-10 | validation error at convergence: 1% gain at 100 samples, up to 3% at 50000 | Improved |
| fix random polarities (RAND-Polarity) (ablation) | AlexNet · Fashion-MNIST and CIFAR-10 | validation error at convergence vs Fluid-Net | No clear effect |
| transfer ImageNet polarity pattern (fixed) instead of ImageNet weights (finetune) | AlexNet · Fashion-MNIST and CIFAR-10, limited data | validation error at convergence: significant for Fashion-MNIST 250, 1000 and CIFAR-10 ≤1000 except 250 | Improved |
| initialize Frozen-Net with ImageNet weight magnitudes (IN-Weight) instead of random magnitudes (ablation) | Frozen-Net AlexNet · Fashion-MNIST and CIFAR-10, limited data | validation error / epochs | No clear effect |
| initialize Frozen-Net with ImageNet weight magnitudes when data and time are abundant (ablation) | Frozen-Net AlexNet · Fashion-MNIST and CIFAR-10, abundant data | validation accuracy: Fashion-MNIST gain ~1%, CIFAR-10 gain ~4% | Improved |
| reduce network size for randomly polarity-fixed Frozen-Net (ablation) | Frozen-Net RAND-Polarity on XOR-5D · XOR-5D, small networks (e.g., 10 hidden units) | probability of learning and sample/time efficiency | Hurt |
| increase network size for randomly polarity-fixed Frozen-Net (ablation) | Frozen-Net RAND-Polarity on XOR-5D · XOR-5D, >15 hidden units | probability of learning XOR-5D: >99% chance at 15 hidden units | Improved |
| Freeze polarities with Freeze-SGD using Adam-based training that resets sign-violating weights | Frozen-Net generally · general | optimization disadvantage | Hurt |
Closest papers by meaning
- How connectivity structure shapes rich and lazy learning in neural circuits (ICLR 2024 · similarity 0.85)
- Dynamics of Supervised and Reinforcement Learning in the Non-Linear Perceptron (NeurIPS 2024 · similarity 0.84)
- A Theory of Initialisation's Impact on Specialisation (ICLR 2025 · similarity 0.84)
- The impact of allocation strategies in subset learning on the expressive power of neural networks (ICLR 2025 · similarity 0.84)
- Rotational Equilibrium: How Weight Decay Balances Learning Across Neural Networks (ICML 2024 · similarity 0.84)
- Strategy Coopetition Explains the Emergence and Transience of In-Context Learning (ICML 2025 · similarity 0.84)
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.
