Research map / Architectures and efficiency / CNN design
Fast Algorithms for Convolutional Neural Networks
Check 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 |
|---|---|---|---|
| Use Winograd F(2x2,3x3) instead of cuDNN convolution | VGG Network E · VGG Network E, fp32, batch sizes 1–64 | Effective TFLOPS: 2.26X; 2.06X; 5.91X; 7.28X; 3.09X; 1.79X; 1.48X | Helped |
| Use Winograd F(2x2,3x3) instead of cuDNN convolution | VGG Network E · VGG Network E, fp16, batch sizes 1–64 | Effective TFLOPS: 2.64X; 2.13X; 5.95X; 7.42X; 3.24X; 1.92X; 1.63X | Helped |
| Use F(2x2,3x3) rather than direct convolution | VGG convolution layers with fp32 data · VGG Network E layers 1.2–5 | maximum element error: 4.01E-05 -> 1.53E-05; 8.01E-05 -> 2.86E-05; 1.53E-04 -> 5.34E-05; 3.20E-04 -> 5.34E-05; 3.43E-04 -> 4.20E-05 | Helped |
| Use F(4x4,3x3) rather than direct convolution | VGG convolution layers · VGG Network E, fp32 | maximum element error: 4.01E-05 -> 2.84E-04; 8.01E-05 -> 5.41E-04; 1.53E-04 -> 9.06E-04; 3.20E-04 -> 1.04E-03; 3.43E-04 -> 1.08E-03 | Hurt |
Closest papers by meaning
- DWM: A Decomposable Winograd Method for Convolution Acceleration (AAAI 2020 · similarity 0.87)
- Deep Tensor Convolution on Multicores (ICML 2017 · similarity 0.87)
- Efficient Sparse-Winograd Convolutional Neural Networks (ICLR 2018 · similarity 0.85)
- Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks (CVPR 2023 · similarity 0.84)
- SBNet: Sparse Blocks Network for Fast Inference (CVPR 2018 · similarity 0.84)
- FlashFFTConv: Efficient Convolutions for Long Sequences with Tensor Cores (ICLR 2024 · similarity 0.83)
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.