Research map / Architectures and efficiency / CNN design
Functional Neural Networks: Shift invariant models for functional data with applications to EEG classification
Florian Heinrichs, Mavin Heim, Corinna Weber
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 |
|---|---|---|---|
| replace scalar layers with functional convolutional and dense layers (FNN) | simulated datasets I and II · simulated data, N=1000 to 5000 | accuracy | Improved |
| use shift-invariant functional convolutions instead of MLP-style layers | sliding-window simulated dataset II · simulated dataset II | accuracy: MLP 79.0%-90.7% vs FNN 99.6%-99.9% | Improved |
| functional networks (FNN) vs EEGNet on phoneme dataset | phoneme periodogram classification · phoneme | accuracy: 91.53 vs 87.83 | Improved |
| FNN vs EEGNet under the classic (cue-aligned) EEG approach | BCI Competition IV 2A classic 4-class · EEG classic approach | accuracy: EEGNet 72.13 vs best FNN 69.08 | Hurt |
| functional CNN (FNN(40)) vs EEGNet under sliding windows | BCI Competition IV 2A sliding-window 7-class · EEG sliding windows | accuracy: 56.29 vs 51.81 (+4.48) | Improved |
| fully functional (per-time-point) predictions vs scalar window classification (ablation) | Func2Func models on sliding-window EEG · EEG sliding windows | accuracy: Func2Func(2) 51.52 and Func2Func(3) 47.33 vs scalar FNN(40) 56.29 | Hurt |
| reduce filters from 40+20 to 5+10 in sliding-window FNN (ablation) | FNN for sliding-window EEG · EEG sliding windows | accuracy: 52.12 vs 51.70 for 40+20 | No clear effect |
| single functional dense layer (F-MLP) instead of functional convolutions on classic EEG (ablation) | classic EEG approach · EEG classic approach | accuracy: F-MLP(20) 66.31 vs FNN(20) 67.23 | No clear effect |
| increase filters from 20 to 40 in the single-conv-layer sliding-window FNN (FNN(20) to FNN(40)) (ablation) | FNN sliding-window EEG · EEG sliding windows | accuracy: 55.43 vs 56.29 | Improved |
Closest papers by meaning
- Functional Connectomes of Neural Networks (AAAI 2025 · similarity 0.83)
- What do EEG Foundation Models Capture from Human Brain Signals? (NeurIPS 2026 · similarity 0.83)
- Decoding Natural Images from EEG for Object Recognition (ICLR 2024 · similarity 0.82)
- NeuroAtlas: Benchmarking Foundation Models for Clinical EEG and Brain-Computer Interfaces (NeurIPS 2026 · similarity 0.82)
- Brain-OF: An Omnifunctional Foundation Model for fMRI, EEG and MEG (NeurIPS 2026 · similarity 0.82)
- BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals (NeurIPS 2025 · 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.
