Research map / Architectures and efficiency / RNNs and normalization
Temporal FiLM: Capturing Long-Range Sequence Dependencies with Feature-Wise Modulations.
Sawyer Birnbaum, Volodymyr Kuleshov, Zayd Enam, Pang Wei W Koh, Stefano Ermon
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 |
|---|---|---|---|
| add TFiLM layers (ablation) | conv model · Yelp-2 | accuracy: 78.1% -> 95.6% | Improved |
| add TFiLM layers (ablation) | conv model · Yelp-5 | accuracy: 61.5% -> 62.3% | Improved |
| add TFiLM layers (ablation) | conv model · Audio SR MultiSpeaker r=8 | SNR (dB): 9.9 -> 12.0 | Improved |
| add TFiLM layers (ablation) | conv model · Audio SR SingleSpeaker r=2 | SNR (dB): 19.4 -> 19.5 | No clear effect |
| add TFiLM layers (ablation) | conv model · Audio SR Piano r=8 | SNR (dB): 11.1 -> 13.3 | Improved |
| add TFiLM layers (ablation) | conv model · Audio SR (all tasks, ratios) | SNR (dB): +1.0 dB on average | Improved |
| add TFiLM layers (ablation) | conv model · Audio SR (all tasks, ratios) | LSD (dB): 0.2 dB on average | Improved |
| add TFiLM layers (ablation) | conv model · ChIP-seq H3K4me3 | Pearson correlation: 0.66 -> 0.90 | Improved |
| add TFiLM layers (ablation) | conv model · ChIP-seq H3K27me3 | Pearson correlation: 0.65 -> 0.64 | No clear effect |
| remove skip connections (ablation) | TFiLM audio model · Audio SR | reconstruction accuracy | Hurt |
| train on speech, test on piano (or vice versa) | TFiLM audio model · Audio SR cross-dataset | performance | Hurt |
Closest papers by meaning
- Long-Term Recurrent Convolutional Networks for Visual Recognition and Description (TPAMI 2016 · similarity 0.86)
- Sequence Modeling with Multiresolution Convolutional Memory (ICML 2023 · similarity 0.86)
- Combining Recurrent, Convolutional, and Continuous-time Models with Linear State Space Layers (NeurIPS 2021 · similarity 0.85)
- Learning Sequence Representations by Non-local Recurrent Neural Memory (IJCV 2022 · similarity 0.85)
- Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods (NeurIPS 2019 · similarity 0.85)
- Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling (NeurIPS 2024 · similarity 0.84)
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.
