Research map / Architectures and efficiency / CNN design
MUXConv: Information Multiplexing in Convolutional Neural Networks
Zhichao Lu, Kalyanmoy Deb, Vishnu Naresh Boddeti
Check 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 |
|---|---|---|---|
| Add spatial multiplexing to depth-wise separable convolutions (ablation) | MobileNetV2 · ImageNet at low-MAdds model sizes | Top-1 accuracy: 5.8% | Improved |
| Add spatial multiplexing to depth-wise separable convolution layers (ablation) | MobileNetV2 · ImageNet at fixed spatial resolution | Top-1 accuracy | Improved |
| Leave out 25% of input channels in normal blocks (ablation) | MobileNetV2 with channel multiplexing · ImageNet | Parameters: 13% | Improved |
| Leave out 25% of input channels in normal blocks (ablation) | MobileNetV2 with channel multiplexing · ImageNet | MAdds: 20% | Improved |
| Leave out 25% of input channels in normal blocks (ablation) | MobileNetV2 with channel multiplexing · ImageNet | Predictive accuracy | No clear effect |
| Solve preference-decomposed subproblems collectively instead of one at a time (ablation) | PBI-based architecture search · NASBench101 | Search effectiveness under fixed budget | Improved |
| Use MUXNet-m instead of MobileNetV2 as the SSDLite backbone | SSDLite · PASCAL VOC 2007 test | mAP: 67.4 -> 68.6 | Improved |
| Use MUXNet-m instead of MobileNetV2 as the PPM backbone | PPM segmentation · ADE20K validation | mIoU: 35.76 -> 35.80 | No clear effect |
| Use MUXNet-m instead of CheXNet | ChestX-Ray14 transfer learning · ChestX-Ray14 test | AUROC: 84.4 -> 84.1 | No clear effect |
| Use MUXNet-m instead of MobileNetV2 | ADE20K segmentation with C1 head · ADE20K validation | mIoU: 34.84 -> 32.42 | Hurt |
| Use MUXNet-m instead of MobileNetV2 | CIFAR transfer learning · CIFAR-10 test | Top-1 accuracy: 95.74 -> 98.00 | Mixed |
Closest papers by meaning
- Bilinear Convolutional Neural Networks for Fine-Grained Visual Recognition (TPAMI 2017 · similarity 0.87)
- SCNN: A General Distribution Based Statistical Convolutional Neural Network with Application to Video Object Detection (AAAI 2019 · similarity 0.87)
- Region-Based Convolutional Networks for Accurate Object Detection and Segmentation (TPAMI 2016 · similarity 0.87)
- MicroNet: Improving Image Recognition With Extremely Low FLOPs (ICCV 2021 · similarity 0.87)
- ESPNetv2: A Light-Weight, Power Efficient, and General Purpose Convolutional Neural Network (CVPR 2019 · similarity 0.86)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 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.
