Research map / Architectures and efficiency / RNNs and normalization
Minimizing FLOPs to Learn Efficient Sparse Representations
Biswajit Paria, Chih-Kuan Yeh, Ian E.H. Yen, Ning Xu, Pradeep Ravikumar, Barnabás Póczos
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 |
|---|---|---|---|
| Use the FLOPs regularizer instead of L1 regularization | 64-dimensional sparse embeddings · CIFAR-100 test | FLOPs-per-row: 0.47 -> 0.40 | Improved |
| Use the FLOPs regularizer instead of L1 regularization | 64-dimensional sparse embeddings · CIFAR-100 test | prec@4: 54.32 -> 55.23 | Improved |
| Use the FLOPs regularizer instead of L1 regularization | 64-dimensional sparse embeddings · CIFAR-100 test | prec@16: 54.96 -> 55.21 | Improved |
| Increase the number of active hash bits from k=1 to k=3 (ablation) | SDH · CIFAR-100 test | prec@4: 57.22 -> 55.81 | Hurt |
| Increase the number of active hash bits from k=1 to k=3 (ablation) | SDH · CIFAR-100 test | FLOPs-per-row: 1.18 -> 8.82 | Hurt |
| Use sparse FLOPs-regularized embeddings instead of dense embeddings | 64-dimensional embeddings · CIFAR-100 test | FLOPs-per-row: 64 -> 0.40 | Improved |
| Use sparse FLOPs-regularized embeddings instead of dense embeddings | 64-dimensional embeddings · CIFAR-100 test | prec@4: 57.31 -> 55.23 | Hurt |
| Use sparse FLOPs-regularized embeddings instead of SDH k=3 | 64-dimensional embeddings · CIFAR-100 test | FLOPs-per-row: 8.82 -> 0.40 | Improved |
| Use sparse FLOPs-regularized embeddings instead of SDH k=3 | 64-dimensional embeddings · CIFAR-100 test | prec@4: 55.81 -> 55.23 | Hurt |
Closest papers by meaning
- Approximate Search with Quantized Sparse Representations (ECCV 2016 · similarity 0.88)
- Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation (ICML 2025 · similarity 0.88)
- Sparse Dimensionality Reduction Revisited (ICML 2024 · similarity 0.86)
- Efficient end-to-end learning for quantizable representations (ICML 2018 · similarity 0.85)
- End-To-End Efficient Representation Learning via Cascading Combinatorial Optimization (CVPR 2019 · similarity 0.85)
- SparseFormer: Sparse Visual Recognition via Limited Latent Tokens (ICLR 2024 · 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.
