Research map / Probabilistic and scientific modeling / Variational and generative sampling
Invertible Residual Networks
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 |
|---|---|---|---|
| Apply the Lipschitz constraint with c = 0.9 instead of removing it (ablation) | i-ResNet classifier · CIFAR10 | classification error (%): 6.69 -> 6.78 | No clear effect |
| Apply the Lipschitz constraint with c = 0.9 instead of removing it (ablation) | i-ResNet classifier · CIFAR100 | classification error (%): 23.97 -> 24.58 | Mixed |
| Use stronger spectral normalization (lower c) (ablation) | i-ResNet classifier · CIFAR10 | classification error (%): 6.78 -> 8.71 | Hurt |
| Use stronger spectral normalization (lower c) (ablation) | i-ResNet classifier · CIFAR100 | classification error (%): 24.58 -> 29.45 | Hurt |
| Use stronger spectral normalization (lower c) (ablation) | i-ResNet classifier · MNIST | classification error (%): 0.40 -> 0.86 | Hurt |
| Use i-ResNet block structure rather than Glow coupling blocks | CIFAR10 classifier · CIFAR10 | classification error: 12.63 -> 8.03 | Helped |
| Remove injective padding (ablation) | i-ResNet CIFAR10 classifier · CIFAR10 | accuracy | Hurt |
| Use i-ResNet flow instead of Glow | Generative model · MNIST | bits/dim: 1.05 -> 1.06 | Hurt |
| Use i-ResNet flow instead of Glow | Generative model · CIFAR10 | bits/dim: 3.35 -> 3.45 | Hurt |
| Use i-ResNet flow instead of FFJORD | Generative model · CIFAR10 | bits/dim: 3.40 -> 3.45 | Hurt |
Closest papers by meaning
- Residual Flows for Invertible Generative Modeling (NeurIPS 2019 · similarity 0.92)
- MintNet: Building Invertible Neural Networks with Masked Convolutions (NeurIPS 2019 · similarity 0.90)
- i-RevNet: Deep Invertible Networks (ICLR 2018 · similarity 0.90)
- The Reversible Residual Network: Backpropagation Without Storing Activations (NeurIPS 2017 · similarity 0.90)
- Invertibility of Convolutional Generative Networks from Partial Measurements (NeurIPS 2018 · similarity 0.89)
- Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers (NeurIPS 2025 · similarity 0.87)
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.