Papers

Field

Recent papers

All papers in Research

arXiv cs.NESep 18
Predictive Suppression Layers for Communication-Efficient Spiking Neural Networks
Aidin Attar, Michele Rossi
Feedforward Spiking Neural Networks (SNNs) typically propagate every generated spike indiscriminately, disregarding whether the information is redundant from an information-theoretic perspective. This lack of selectivity induces high redundancy in inter-layer communication, creating an expensive overhead, e.g., in scenarios involving many-core neuromorphic hardware or communication-dominated Internet-of-Things (IoT) where features are transmitted wirelessly. To address this challenge, we trade localized processing for leaner network channels by introducing a minimal predictive coding framework for SNNs. We propose two layer variants sharing a predictor block: error units, which transmit signed spiking residuals, and predictive suppression, which uses residual magnitude to dynamically gate and forward only unpredictable, "surprising" activity. Evaluated on the N-MNIST and Spiking Heidelberg Digits (SHD) datasets using diagnostic metrics that decouple local processing from cross-layer communication, our new predictive coding layers achieve significant communication savings. Numerical results reveal a three-fold reduction in communicated activity, while increasing the task accuracy…
PDF arXiv
arXiv cs.LGSep 18
Efficient Architecture Search under Leave-One-Subject-Out Evaluation
Heinke Hihn, Friedhelm Schwenker
Deep neural architectures are widely used for signal processing in automated pain assessment systems. However, architecture design has remained largely a manual task despite the potential efficiency benefits of Neural Architecture Search (NAS). Embedding NAS in a Leave-One-Subject-Out (LOSO) evaluation is computationally demanding because a fully nested implementation requires $N$ independent architecture searches and, assuming approximately linear training cost, scales as $\mathcal{O}(N^2)$. We propose a block-based, leakage-controlled approach that shares NAS runs between subjects, reducing the number of searches from $N$ to $B$, where $B \ll N$, dubbed PainNAS. On the BioVid Heat Pain dataset, PainNAS yields comparable subject-level accuracy with substantially fewer parameters and FLOPs.
PDF arXiv

Papers

Topics in this field

eeg foundation models · electroencephalography eeg · eeg foundation model132 papers15 this week
electrocardiogram ecg · electrocardiogram · ecgs130 papers9 this week
spiking neural networks · networks spiking neural · spiking neural network50 papers6 this week

Recent papers

All papers in Research

arXiv cs.NESep 18
Predictive Suppression Layers for Communication-Efficient Spiking Neural Networks
Aidin Attar, Michele Rossi
Feedforward Spiking Neural Networks (SNNs) typically propagate every generated spike indiscriminately, disregarding whether the information is redundant from an information-theoretic perspective. This lack of selectivity induces high redundancy in inter-layer communication, creating an expensive overhead, e.g., in scenarios involving many-core neuromorphic hardware or communication-dominated Internet-of-Things (IoT) where features are transmitted wirelessly. To address this challenge, we trade localized processing for leaner network channels by introducing a minimal predictive coding framework for SNNs. We propose two layer variants sharing a predictor block: error units, which transmit signed spiking residuals, and predictive suppression, which uses residual magnitude to dynamically gate and forward only unpredictable, "surprising" activity. Evaluated on the N-MNIST and Spiking Heidelberg Digits (SHD) datasets using diagnostic metrics that decouple local processing from cross-layer communication, our new predictive coding layers achieve significant communication savings. Numerical results reveal a three-fold reduction in communicated activity, while increasing the task accuracy…
PDF arXiv
arXiv cs.LGSep 18
Efficient Architecture Search under Leave-One-Subject-Out Evaluation
Heinke Hihn, Friedhelm Schwenker
Deep neural architectures are widely used for signal processing in automated pain assessment systems. However, architecture design has remained largely a manual task despite the potential efficiency benefits of Neural Architecture Search (NAS). Embedding NAS in a Leave-One-Subject-Out (LOSO) evaluation is computationally demanding because a fully nested implementation requires $N$ independent architecture searches and, assuming approximately linear training cost, scales as $\mathcal{O}(N^2)$. We propose a block-based, leakage-controlled approach that shares NAS runs between subjects, reducing the number of searches from $N$ to $B$, where $B \ll N$, dubbed PainNAS. On the BioVid Heat Pain dataset, PainNAS yields comparable subject-level accuracy with substantially fewer parameters and FLOPs.
PDF arXiv

Active contributors · 90 days

Chenyu LiuDEFT: Disentanglement-Enhanced Fine-Tuning for EEG Foundation Models5
Muhammad ShafiqueThinkNet: Evolutionary Deep Ensembles for Efficient Motor-Imagery EEG Classification5
Cuntai GuanDo Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling4
Constantino Álvarez CasadoLearning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition4
Le NguyenLearning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition4
Manuel Lage CañellasLearning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition4

Activity · 12 weeks

Fields · AI · updated 12m ago

Electrocardiogram ecg

Found by the topic model: electrocardiogram ecg, eeg foundation models, spiking neural networks.

30papers this week−28.6% vs last week

All papers · 312 With code · 0

Newest Most relevant

226 papers this month

Open problems in Electrocardiogram ecg

Lack of robust cross-subject generalization in EEG models2 papers raise it · 1 addresses it
Limits real-world deployment of EEG-based systems due to intersubject variability and nonstationarity.
- Controlled Evaluation of EEG Foundation Models: Random Initialisation, Band-Power and Validation-Selected Probesraises
- Memory-Augmented Online Test-Time Adaptation for EEG Motor Imagery via Perturbation Consistency and Confidence Calibrationraises · addresses
Insufficient validation controls for EEG foundation models1 paper raises it · 1 addresses it
Leads to overestimated performance and unclear benefits of pretraining.
- Controlled Evaluation of EEG Foundation Models: Random Initialisation, Band-Power and Validation-Selected Probesraises · addresses
Motion artifact robustness in wrist PPG heart rate estimation1 paper raises it · 1 addresses it
Affects accuracy and reliability of wearable heart rate monitoring.
- An Agent Architecture with Tools and Memory for Selective Correction of Wrist PPG Heart Rate Estimatesraises · addresses
High bandwidth requirements for physiological signal transmission1 paper raises it · 1 addresses it
Restricts scalability and real-time applicability of wearable systems.
- EMG Tokenization Preserves Movement Information at a Fraction of the Bitrateraises · addresses
Limited interpretability of EEG-to-language models1 paper raises it · 1 addresses it
Makes it hard to verify if models truly understand EEG signals rather than generating plausible but incorrect outputs.
- EEG-Captioner: Grounding EEG-to-Language Generation in Fundamental Signal Propertiesraises · addresses

Named by an AI from these papers' abstracts.

Field

Papers

Field

Recent papers

Predictive Suppression Layers for Communication-Efficient Spiking Neural Networks

Efficient Architecture Search under Leave-One-Subject-Out Evaluation

Papers

Topics in this field

Recent papers

Predictive Suppression Layers for Communication-Efficient Spiking Neural Networks

Efficient Architecture Search under Leave-One-Subject-Out Evaluation

Active contributors · 90 days

Activity · 12 weeks

Electrocardiogram ecg

Open problems in Electrocardiogram ecg