Research map / Architectures and efficiency / RNNs and normalization
Deep Complex Networks
Chiheb Trabelsi, Olexa Bilaniuk, Ying Zhang, Dmitriy Serdyuk, Sandeep Subramanian, Joao Felipe Santos, Soroush Mehri, Negar Rostamzadeh, Yoshua Bengio, Christopher J Pal
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 a complex-valued network with CReLU instead of a real-valued network | wide-and-shallow residual network · CIFAR-100 | classification error: 27.22 -> 26.36 | Improved |
| use a complex-valued network with CReLU instead of a real-valued network | wide-and-shallow residual network · CIFAR-10 | classification error: 5.42 -> 6.17 | Hurt |
| use a complex-valued network with CReLU instead of a real-valued network | wide-and-shallow residual network · SVHN* | classification error: 3.42 -> 3.70 | Hurt |
| use the deep complex model instead of the deep real model | MusicNet transcription · MusicNet | average precision: 69.6 -> 72.9 | Improved |
| use a complex convolutional LSTM instead of a real convolutional LSTM | speech spectrum prediction · TIMIT test | MSE: 12.18 -> 11.90 | Improved |
| replace CReLU with zReLU (ablation) | complex residual networks · CIFAR-10 | classification error: 6.17 -> 11.71 | Hurt |
| replace CReLU with modReLU (ablation) | complex residual networks · CIFAR-10 | classification error: 6.17 -> 23.42 | Hurt |
| replace CReLU with zReLU (ablation) | complex residual networks · SVHN* | classification error: 3.70 -> 80.41 | Hurt |
| replace complex batch normalization with naive complex batch normalization (ablation) | complex residual networks · SVHN* | classification error | Hurt |
Closest papers by meaning
- Toward Complex-Valued Neural Networks for Waveform Generation (ICLR 2026 · similarity 0.86)
- Bayesian Sparsification of Deep C-valued Networks (ICML 2020 · similarity 0.86)
- FCCNs: Fully Complex-valued Convolutional Networks using Complex-valued Color Model and Loss Function (ICCV 2023 · similarity 0.86)
- Optimal approximation using complex-valued neural networks (NeurIPS 2023 · similarity 0.85)
- Real-Valued Backpropagation is Unsuitable for Complex-Valued Neural Networks (NeurIPS 2022 · similarity 0.84)
- Which Frequencies do CNNs Need? Emergent Bottleneck Structure in Feature Learning (ICML 2024 · similarity 0.82)
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.
