Research map / Architectures and efficiency / RNNs and normalization
Understanding Hyperdimensional Computing for Parallel Single-Pass Learning
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 |
|---|---|---|---|
| Initialize basis hypervectors with RFF-based target similarities instead of classical HDC initialization | HDC · MNIST | accuracy (10-Epo, %): 89.0 -> 95.4; 6.4% | Helped |
| Use cyclic group VSA elements of order 2^3 instead of binary elements (ablation) | RFF HDC · MNIST | accuracy (10-Epo, %): 95.4 -> 95.7; 1.2% | Helped |
| Use cyclic group VSA elements of order 2^4 instead of binary elements (ablation) | RFF HDC · MNIST | accuracy (10-Epo, %): 95.4 -> 96.6 | Helped |
| Increase cyclic group order from 2^3 to 2^4 (ablation) | RFF group VSA · ISOLET | accuracy (10-Epo, %): 94.4 -> 96.0 | Helped |
| Increase cyclic group order from 2^3 to 2^4 (ablation) | RFF group VSA · UCIHAR | accuracy (10-Epo, %): 95.6 -> 96.6 | Helped |
| Increase cyclic group order from 2^3 to 2^4 (ablation) | RFF group VSA · Fashion-MNIST | accuracy (10-Epo, %): 86.7 -> 86.5 | Hurt |
| Train for 10 epochs instead of one (ablation) | RFF HDC · MNIST | accuracy (%): 95.4 -> 95.4 | No clear effect |
| Train for 10 epochs instead of one (ablation) | RFF G(2^3)-VSA · Fashion-MNIST | accuracy (%): 85.4 -> 86.7 | Helped |
| Train for 10 epochs instead of one (ablation) | RFF G(2^4)-VSA · Fashion-MNIST | accuracy (%): 87.4 -> 86.5 | Hurt |
| Use RFF HDC instead of the perceptron baseline | ISOLET · 10,000-dimensional models | accuracy (1-Epo, %): 82.8 -> 90.6 | Helped |
| Use group VSA of order 2^3 instead of the 1-bit RFF perceptron | Fashion-MNIST · 10,000-dimensional models | accuracy (10-Epo, %): 79.5 -> 86.7 | Helped |
| Increase group order from 2^3 to 2^4 (ablation) | RFF group VSA · MNIST, 1k-dimensional hypervectors | accuracy (10-Epo, %): 88.61% -> 92.56% | Helped |
Closest papers by meaning
- Bridging the Gap Between Hyperdimensional Computing and Kernel Methods via the Nyström Method (AAAI 2025 · similarity 0.89)
- G-Net: A Provably Easy Construction of High-Accuracy Random Binary Neural Networks (NeurIPS 2025 · similarity 0.85)
- HDQMF: Holographic Feature Decomposition Using Quantum Algorithms (CVPR 2024 · similarity 0.83)
- LibAMM: Empirical Insights into Approximate Computing for Accelerating Matrix Multiplication (NeurIPS 2024 · similarity 0.82)
- Separations in the Representational Capabilities of Transformers and Recurrent Architectures (NeurIPS 2024 · similarity 0.81)
- Scaling Probabilistic Circuits via Monarch Matrices (ICML 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.