David Atienza

William Andrew Simon, Una Pale, Tomas Teijeiro et al.

HyperDimensional Computing (HDC) as a machine learning paradigm is highly interesting for applications involving continuous, semi-supervised learning for long-term monitoring. However, its accuracy is not yet on par with other Machine Learning (ML) approaches. Frameworks enabling fast design space exploration to find practical algorithms are necessary to make HD computing competitive with other ML techniques. To this end, we introduce HDTorch, an open-source, PyTorch-based HDC library with CUDA extensions for hypervector operations. We demonstrate HDTorch's utility by analyzing four HDC benchmark datasets in terms of accuracy, runtime, and memory consumption, utilizing both classical and online HD training methodologies. We demonstrate average (training)/inference speedups of (111x/68x)/87x for classical/online HD, respectively. Moreover, we analyze the effects of varying hyperparameters on runtime and accuracy. Finally, we demonstrate how HDTorch enables exploration of HDC strategies applied to large, real-world datasets. We perform the first-ever HD training and inference analysis of the entirety of the CHB-MIT EEG epilepsy database. Results show that the typical approach of training on a subset of the data does not necessarily generalize to the entire dataset, an important factor when developing future HD models for medical wearable devices.

Denisa-Andreea Constantinescu, David Atienza

At global scale, data-center electricity demand is growing faster than the grids that supply it, while system operators increasingly require large flexible loads that can adjust power within seconds to absorb variable wind and solar generation. For multi-megawatt AI/HPC facilities, the key unresolved question is practical and measurable: how quickly can the software stack translate a grid request into a real change in GPU power at the facility meter, where commitments are settled? We answer this on real hardware with GridPilot, a three-tier predictive controller operating across milliseconds, seconds, and hours, augmented by a deterministic safety-island bypass for fast response. On a three-GPU NVIDIA V100 testbed, GridPilot achieves a measured end-to-end trigger-to-target response of 97.2 ms, which is 6.9x faster than the 700 ms requirement of Nordic Fast Frequency Reserve. We further incorporate an instantaneous Power Usage Effectiveness (PUE) correction so dispatched commitments remain robust at meter level rather than only at IT load level. In replay experiments across six representative European grids (from Sweden to Poland), the PUE-aware controller closes 2.5-5.8 percentage points of cooling-overhead drag. GridPilot is released as open source and serves as a proof of concept that MW-scale AI/HPC demand can be engineered as controllable, grid-responsive flexibility by design.

Simone Machetti, Pasquale Davide Schiavone, Lara Orlandic et al.

Graphics processing units (GPUs) excel at parallel processing, but remain largely unexplored in ultra-low-power edge devices (TinyAI) due to their power and area limitations, as well as the lack of suitable programming frameworks. To address these challenges, this work introduces embedded GPU (e-GPU), an open-source and configurable RISC-V GPU platform designed for TinyAI devices. Its extensive configurability enables area and power optimization, while a dedicated Tiny-OpenCL implementation provides a lightweight programming framework tailored to resource-constrained environments. To demonstrate its adaptability in real-world scenarios, we integrate the e-GPU with the eXtendible Heterogeneous Energy-Efficient Platform (X-HEEP) to realize an accelerated processing unit (APU) for TinyAI applications. Multiple instances of the proposed system, featuring varying e-GPU configurations, are implemented in TSMC's 16 nm SVT CMOS technology and are operated at 300 MHz and 0.8 V. Their area and leakage characteristics are analyzed to ensure alignment with TinyAI constraints. To assess both runtime overheads and computational efficiency, we employ two benchmarks: General Matrix Multiply (GeMM) and bio-signal processing (TinyBio) workloads. The GeMM benchmark is used to quantify the scheduling overhead introduced by the Tiny-OpenCL framework. The results show that the delay becomes negligible for matrix sizes larger than 256x256 (or equivalent problem sizes). The TinyBio benchmark is then used to evaluate performance and energy improvements over the baseline host under pure processing conditions. The results indicate that the high-range e-GPU configuration with 16 threads achieves up to a 15.1x speed-up and reduces energy consumption by up to 3.1x, while incurring only a 2.5x area overhead and operating within a 28 mW power budget.

Saleh Baghersalimi, Alireza Amirshahi, Farnaz Forooghifar et al.

Integrating low-power wearable Internet of Things (IoT) systems into routine health monitoring is an ongoing challenge. Recent advances in the computation capabilities of wearables make it possible to target complex scenarios by exploiting multiple biosignals and using high-performance algorithms, such as Deep Neural Networks (DNNs). There is, however, a trade-off between performance of the algorithms and the low-power requirements of IoT platforms with limited resources. Besides, physically larger and multi-biosignal-based wearables bring significant discomfort to the patients. Consequently, reducing power consumption and discomfort is necessary for patients to use IoT devices continuously during everyday life. To overcome these challenges, in the context of epileptic seizure detection, we propose a many-to-one signals knowledge distillation approach targeting single-biosignal processing in IoT wearable systems. The starting point is to get a highly-accurate multi-biosignal DNN, then apply our approach to develop a single-biosignal DNN solution for IoT systems that achieves an accuracy comparable to the original multi-biosignal DNN. To assess the practicality of our approach to real-life scenarios, we perform a comprehensive simulation experiment analysis on several state-of-the-art edge computing platforms, such as Kendryte K210 and Raspberry Pi Zero.

Una Pale, Tomas Teijeiro, David Atienza

The healthcare landscape is moving from the reactive interventions focused on symptoms treatment to a more proactive prevention, from one-size-fits-all to personalized medicine, and from centralized to distributed paradigms. Wearable IoT devices and novel algorithms for continuous monitoring are essential components of this transition. Hyperdimensional (HD) computing is an emerging ML paradigm inspired by neuroscience research with various aspects interesting for IoT devices and biomedical applications. Here we explore the not yet addressed topic of optimal encoding of spatio-temporal data, such as electroencephalogram (EEG) signals, and all information it entails to the HD vectors. Further, we demonstrate how the HD computing framework can be used to perform feature selection by choosing an adequate encoding. To the best of our knowledge, this is the first approach to performing feature selection using HD computing in the literature. As a result, we believe it can support the ML community to further foster the research in multiple directions related to feature and channel selection, as well as model interpretability.

Hojjat Karami, David Atienza, Jean-Philippe Thiran et al.

Feature engineering for Electronic Health Records (EHR) is complicated by irregular observation intervals, variable measurement frequencies, and structural sparsity inherent to clinical time series. Existing automated methods either lack clinical domain awareness or assume clean, regularly sampled inputs, limiting their applicability to real-world EHR data. We present \textbf{FeatEHR-LLM}, a framework that leverages Large Language Models (LLMs) to generate clinically meaningful tabular features from irregularly sampled EHR time series. To limit patient privacy exposure, the LLM operates exclusively on dataset schemas and task descriptions rather than raw patient records. A tool-augmented generation mechanism equips the LLM with specialized routines for querying irregular temporal data, enabling it to produce executable feature-extraction code that explicitly handles uneven observation patterns and informative sparsity. FeatEHR-LLM supports both univariate and multivariate feature generation through an iterative, validation-in-the-loop pipeline. Evaluated on eight clinical prediction tasks across four ICU datasets, our framework achieves the highest mean AUROC on 7 out of 8 tasks, with improvements of up to 6 percentage points over strong baselines. Code is available at github.com/hojjatkarami/FeatEHR-LLM.

Lara Orlandic, Tomas Teijeiro, David Atienza

Cough audio signal classification is a potentially useful tool in screening for respiratory disorders, such as COVID-19. Since it is dangerous to collect data from patients with such contagious diseases, many research teams have turned to crowdsourcing to quickly gather cough sound data, as it was done to generate the COUGHVID dataset. The COUGHVID dataset enlisted expert physicians to diagnose the underlying diseases present in a limited number of uploaded recordings. However, this approach suffers from potential mislabeling of the coughs, as well as notable disagreement between experts. In this work, we use a semi-supervised learning (SSL) approach to improve the labeling consistency of the COUGHVID dataset and the robustness of COVID-19 versus healthy cough sound classification. First, we leverage existing SSL expert knowledge aggregation techniques to overcome the labeling inconsistencies and sparsity in the dataset. Next, our SSL approach is used to identify a subsample of re-labeled COUGHVID audio samples that can be used to train or augment future cough classification models. The consistency of the re-labeled data is demonstrated in that it exhibits a high degree of class separability, 3x higher than that of the user-labeled data, despite the expert label inconsistency present in the original dataset. Furthermore, the spectral differences in the user-labeled audio segments are amplified in the re-labeled data, resulting in significantly different power spectral densities between healthy and COVID-19 coughs, which demonstrates both the increased consistency of the new dataset and its explainability from an acoustic perspective. Finally, we demonstrate how the re-labeled dataset can be used to train a cough classifier. This SSL approach can be used to combine the medical knowledge of several experts to improve the database consistency for any diagnostic classification task.

Jingwei Sun, Zhixu Du, Anna Dai et al.

Vertical federated learning (VFL) enables a service provider (i.e., active party) who owns labeled features to collaborate with passive parties who possess auxiliary features to improve model performance. Existing VFL approaches, however, have two major vulnerabilities when passive parties unexpectedly quit in the deployment phase of VFL - severe performance degradation and intellectual property (IP) leakage of the active party's labels. In this paper, we propose \textbf{Party-wise Dropout} to improve the VFL model's robustness against the unexpected exit of passive parties and a defense method called \textbf{DIMIP} to protect the active party's IP in the deployment phase. We evaluate our proposed methods on multiple datasets against different inference attacks. The results show that Party-wise Dropout effectively maintains model performance after the passive party quits, and DIMIP successfully disguises label information from the passive party's feature extractor, thereby mitigating IP leakage.

Una Pale, Tomas Teijeiro, David Atienza

Epilepsy is a chronic neurological disorder with a significant prevalence. However, there is still no adequate technological support to enable epilepsy detection and continuous outpatient monitoring in everyday life. Hyperdimensional (HD) computing is an interesting alternative for wearable devices, characterized by a much simpler learning process and also lower memory requirements. In this work, we demonstrate a few additional aspects in which HD computing, and the way its models are built and stored, can be used for further understanding, comparing, and creating more advanced machine learning models for epilepsy detection. These possibilities are not feasible with other state-of-the-art models, such as random forests or neural networks. We compare inter-subject similarity of models per different classes (seizure and non-seizure), then study the process of creation of generalized models from personalized ones, and in the end, how to combine personalized and generalized models to create hybrid models. This results in improved epilepsy detection performance. We also tested knowledge transfer between models created on two different datasets. Finally, all those examples could be highly interesting not only from an engineering perspective to create better models for wearables, but also from a neurological perspective to better understand individual epilepsy patterns.

Alireza Amirshahi, Maedeh H. Toosi, Siamak Mohammadi et al.

Wearable systems provide continuous health monitoring and can lead to early detection of potential health issues. However, the lifecycle of wearable systems faces several challenges. First, effective model training for new wearable devices requires substantial labeled data from various subjects collected directly by the wearable. Second, subsequent model updates require further extensive labeled data for retraining. Finally, frequent model updating on the wearable device can decrease the battery life in long-term data monitoring. Addressing these challenges, in this paper, we propose MetaWearS, a meta-learning method to reduce the amount of initial data collection required. Moreover, our approach incorporates a prototypical updating mechanism, simplifying the update process by modifying the class prototype rather than retraining the entire model. We explore the performance of MetaWearS in two case studies, namely, the detection of epileptic seizures and the detection of atrial fibrillation. We show that by fine-tuning with just a few samples, we achieve 70% and 82% AUC for the detection of epileptic seizures and the detection of atrial fibrillation, respectively. Compared to a conventional approach, our proposed method performs better with up to 45% AUC. Furthermore, updating the model with only 16 minutes of additional labeled data increases the AUC by up to 5.3%. Finally, MetaWearS reduces the energy consumption for model updates by 456x and 418x for epileptic seizure and AF detection, respectively.

Una Pale, Tomas Teijeiro, David Atienza

Epilepsy is a chronic neurological disorder that affects a significant portion of the human population and imposes serious risks in the daily life of patients. Despite advances in machine learning and IoT, small, nonstigmatizing wearable devices for continuous monitoring and detection in outpatient environments are not yet available. Part of the reason is the complexity of epilepsy itself, including highly imbalanced data, multimodal nature, and very subject-specific signatures. However, another problem is the heterogeneity of methodological approaches in research, leading to slower progress, difficulty comparing results, and low reproducibility. Therefore, this article identifies a wide range of methodological decisions that must be made and reported when training and evaluating the performance of epilepsy detection systems. We characterize the influence of individual choices using a typical ensemble random-forest model and the publicly available CHB-MIT database, providing a broader picture of each decision and giving good-practice recommendations, based on our experience, where possible.

Yuxuan Wang, María José Belda, Fernando Castro et al.

Modern computing workloads commonly involve matrix-matrix multiplication (mmul) as a core computing pattern. Coarse-Grained Reconfigurable Arrays (CGRAs) can flexibly and efficiently support it, since they combine operation-level reconfigurability and high energy efficiency. However, mapping computational kernels that include mmul with state-of-the-art compilation strategies often leads to suboptimal results, since its multi-dimensional structure hampers the uncovering of its inherent parallelism and, ultimately, runtime performance. Here, we take a different position: we introduce a specialized mmul CGRA kernel schedule, parametrizable across different CGRA sizes. Then, we describe a novel compilation methodology that adapts program representations to effectively leverage it, employing polyhedral transformations to analyze complex computational patterns and expose hidden mmul operations through loop reordering and splitting. The identified patterns are then substituted with optimized assembly, while the remaining program sections are compiled independently. CGRA configurations are then generated, encompassing pre-compiled and compiled parts. Our strategy maximizes resource utilization and ultimately run-time performance, even when mmul is not directly apparent in the source code. The experimental results show speedups up to 9.1x across different benchmarks that contain hidden mmuls and CGRA instances of various sizes.

Jonathan Dan, Una Pale, Alireza Amirshahi et al.

The need for high-quality automated seizure detection algorithms based on electroencephalography (EEG) becomes ever more pressing with the increasing use of ambulatory and long-term EEG monitoring. Heterogeneity in validation methods of these algorithms influences the reported results and makes comprehensive evaluation and comparison challenging. This heterogeneity concerns in particular the choice of datasets, evaluation methodologies, and performance metrics. In this paper, we propose a unified framework designed to establish standardization in the validation of EEG-based seizure detection algorithms. Based on existing guidelines and recommendations, the framework introduces a set of recommendations and standards related to datasets, file formats, EEG data input content, seizure annotation input and output, cross-validation strategies, and performance metrics. We also propose the 10-20 seizure detection benchmark, a machine-learning benchmark based on public datasets converted to a standardized format. This benchmark defines the machine-learning task as well as reporting metrics. We illustrate the use of the benchmark by evaluating a set of existing seizure detection algorithms. The SzCORE (Seizure Community Open-source Research Evaluation) framework and benchmark are made publicly available along with an open-source software library to facilitate research use, while enabling rigorous evaluation of the clinical significance of the algorithms, fostering a collective effort to more optimally detect seizures to improve the lives of people with epilepsy.

Qunyou Liu, Pengbo Yu, Marina Zapater et al.

Deep neural networks (DNNs) are essential for performing advanced tasks on edge or mobile devices, yet their deployment is often hindered by severe resource constraints, including limited memory, energy, and computational power. While uniform quantization provides a straightforward approach to compress model and reduce hardware requirement, it fails to fully leverage the varying robustness across layers, and often lead to accuracy degradation or suboptimal resource usage, particularly at low bitwidths. In contrast, heterogeneous quantization, which allocates different bitwidths to individual layers, can mitigate these drawbacks. Nonetheless, current heterogeneous quantization methods either needs huge brute-force design space search or lacks the adaptability to meet different hardware conditions, such as memory size, energy budget, and latency requirement. Filling these gaps, this work introduces \textbf{\textit{SigmaQuant}}, an adaptive layer-wise heterogeneous quantization framework designed to efficiently balance accuracy and resource usage for varied edge environments without exhaustive search.

Denisa-Andreea Constantinescu, Rubén Rodríguez Álvarez, Jacques Morin et al.

The Square Kilometre Array (SKA) will operate one of the world's largest continuous scientific data systems, sustaining petascale imaging under strict power envelopes. Current radio-interferometric pipelines typically achieve only 4--14\% of hardware peak utilization due to memory and I/O bottlenecks, incurring high energy, operational, and carbon costs, further compounded by the absence of standardised cross-layer metrics and fidelity tolerances for principled hardware--software co-design. We present astroCAMP, a reproducible benchmarking and co-design framework for SKA-scale imaging, contributing: (1) a unified metric suite spanning performance, utilisation, memory/data-movement, sustainability, economics, and scientific fidelity; (2) standardised SKA-representative datasets and benchmark configurations for reproducible cross-platform evaluation; (3) a multi-objective co-design formulation linking quality constraints to time-, energy-, carbon-, and cost-to-solution; and (4) a design-space exploration workflow to derive Pareto-optimal operating regions. We evaluate WSClean+IDG on an AMD EPYC 9334 CPU and NVIDIA H100 GPU, revealing orchestration and synchronization bottlenecks despite efficient kernels, limited CPU strong scaling, and location-dependent carbon/cost efficiency. We illustrate astroCAMP for heterogeneous CPU--FPGA exploration and call on the SKA community to define quantifiable fidelity thresholds to accelerate principled optimisation for SKA-scale imaging.

David Mallasén, Pasquale Davide Schiavone, Alberto A. Del Barrio et al.

Wearable edge AI biomedical devices are increasingly being used for continuous patient health monitoring, enabling real-time insights and extended data collection without the need for prolonged hospital stays. These devices must be energy efficient to minimize battery size, improve comfort, and reduce recharging intervals. This paper investigates the use of specialized low-precision arithmetic formats to enhance the energy efficiency of edge AI biomedical wearables. Specifically, we explore posit arithmetic, a floating-point-like representation, in two biomedical applications that leverage supervised and unsupervised learning algorithms: cough detection for chronic cough monitoring and R peak detection in ECG analysis. Our results reveal that 16-bit posits can replace 32-bit IEEE 754 floating point numbers with minimal accuracy loss in cough detection. For R peak detection, posit arithmetic achieves satisfactory accuracy with as few as 10 or 8 bits, compared to the 16-bit requirement for floating-point formats. To validate these findings beyond algorithm-level simulations, we introduce PHEE, a modular and extensible architecture that integrates the Coprosit posit coprocessor within a RISC-V-based system. Using the X-HEEP framework, PHEE serves as a proof-of-concept platform to quantify the practical energy benefits of low-precision posits in edge AI systems. Post-synthesis results targeting 16 nm TSMC technology show that the posit hardware targeting these ML-based biomedical applications can be 38% smaller and consume up to 42.3% less power at the functional unit level, with no performance compromise. These findings establish the potential of low-precision posit arithmetic to significantly improve the energy efficiency of edge AI biomedical devices.

Christodoulos Kechris, Jonathan Dan, Jose Miranda et al.

Non-linear activation functions are crucial in Convolutional Neural Networks. However, until now they have not been well described in the frequency domain. In this work, we study the spectral behavior of ReLU, a popular activation function. We use the ReLU's Taylor expansion to derive its frequency domain behavior. We demonstrate that ReLU introduces higher frequency oscillations in the signal and a constant DC component. Furthermore, we investigate the importance of this DC component, where we demonstrate that it helps the model extract meaningful features related to the input frequency content. We accompany our theoretical derivations with experiments and real-world examples. First, we numerically validate our frequency response model. Then we observe ReLU's spectral behavior on two example models and a real-world one. Finally, we experimentally investigate the role of the DC component introduced by ReLU in the CNN's representations. Our results indicate that the DC helps to converge to a weight configuration that is close to the initial random weights.

Elena Espinosa, Rubén Rodríguez Álvarez, José Miranda et al.

The advent of next-generation sequencing (NGS) has revolutionized genomic research by enabling cost-effective, high-throughput sequencing of a diverse range of organisms. This breakthrough has unleashed a "Cambrian explosion" in genomic data volume and diversity. This volume of workloads places genomics among the top four big data challenges anticipated for this decade. In this context, pairwise sequence alignment represents a very time- and energy-intensive step in common bioinformatics pipelines. Speeding up this step requires the implementation of heuristic approaches, optimized algorithms, and/or hardware acceleration. Although state-of-the-art CPU and GPU implementations have demonstrated significant performance gains, recent FPGA implementations have shown improved energy efficiency. However, the latter often suffer from limited read-length scalability due to constraints on hardware resources when aligning longer sequences. In this work, we present a flexible FPGA-based accelerator template scalable up to 1000 bp that implements Myers's algorithm to compute exact unit-cost edit-distance using high-level synthesis and a worker-based architecture. GeneTEK, a set of instances of this accelerator template in a Xilinx Zynq UltraScale+ FPGA, achieves up to 113% increase in execution speed and up to 111x reduction in energy consumption compared to leading CPU and GPU solutions, while fitting comparison matrices up to 13x larger than previous FPGA-based systolic-array solutions. By following a SW-HW co-design approach, GeneTEK exploits parallelization at multiple levels and efficient memory use to deliver a scalable and accurate FPGA-based accelerator. These results reaffirm the potential of FPGAs as an energy-efficient platform for pairwise alignment of read-lengths up to 1000 bp.

Christodoulos Kechris, Jonathan Dan, David Atienza

Traditional saliency map methods, popularized in computer vision, highlight individual points (pixels) of the input that contribute the most to the model's output. However, in time-series they offer limited insights as semantically meaningful features are often found in other domains. We introduce Cross-domain Integrated Gradients, a generalization of Integrated Gradients. Our method enables feature attributions on any domain that can be formulated as an invertible, differentiable transformation of the time domain. Crucially, our derivation extends the original Integrated Gradients into the complex domain, enabling frequency-based attributions. We provide the necessary theoretical guarantees, namely, path independence and completeness. Our approach reveals interpretable, problem-specific attributions that time-domain methods cannot capture, on three real-world tasks: wearable sensor heart rate extraction, electroencephalography-based seizure detection, and zero-shot time-series forecasting. We release an open-source Tensorflow/PyTorch library to enable plug-and-play cross-domain explainability for time-series models. These results demonstrate the ability of cross-domain integrated gradients to provide semantically meaningful insights in time-series models that are impossible with traditional time-domain saliency.

Qunyou Liu, Marina Zapater, David Atienza

Transformers have revolutionized AI in natural language processing and computer vision, but their large computation and memory demands pose major challenges for hardware acceleration. In practice, end-to-end throughput is often limited by paged data movement and interconnect bandwidth rather than raw MAC count. This work proposes a unified system-accelerator co-design approach for transformer inference that jointly optimizes a matrix accelerator and its system integration through paged streaming dataflows and explicit overlap of compute and transfer. On the hardware side, we introduce MatrixFlow, a loosely coupled 16x16 systolic-array accelerator with a page-aligned block matrix multiplication method using 4 KB tiles, a small on-chip buffer of about 20 KB, and a pipelined schedule of DMA, compute, and DMA-out to utilize interconnect bandwidth efficiently. On the system side, we develop Gem5-AcceSys, an extension of the gem5 full-system simulator that explores standard interconnects such as PCIe and configurable memory hierarchies including Direct Memory, Direct Cache, and Device Memory modes with SMMU/TLB effects. We evaluate the co-design using gem5 simulations on representative transformer models including BERT and ViT across multiple data types and system setups. Results show up to 22x end-to-end speedup over a CPU-only baseline and 5x to 8x gains over state-of-the-art loosely and tightly coupled accelerators. We further show that a standard PCIe-based host-memory design can achieve about 80 percent of the performance of on-device HBM. Overall, paged streaming and pipeline overlap, rather than large local SRAMs, are the most effective levers for efficient transformer inference under realistic system constraints.

Christodoulos Kechris, Jonathan Dan, Jose Miranda et al.

Deep learning time-series processing often relies on convolutional neural networks with overlapping windows. This overlap allows the network to produce an output faster than the window length. However, it introduces additional computations. This work explores the potential to optimize computational efficiency during inference by exploiting convolution's shift-invariance properties to skip the calculation of layer activations between successive overlapping windows. Although convolutions are shift-invariant, zero-padding and pooling operations, widely used in such networks, are not efficient and complicate efficient streaming inference. We introduce StreamiNNC, a strategy to deploy Convolutional Neural Networks for online streaming inference. We explore the adverse effects of zero padding and pooling on the accuracy of streaming inference, deriving theoretical error upper bounds for pooling during streaming. We address these limitations by proposing signal padding and pooling alignment and provide guidelines for designing and deploying models for StreamiNNC. We validate our method in simulated data and on three real-world biomedical signal processing applications. StreamiNNC achieves a low deviation between streaming output and normal inference for all three networks (2.03 - 3.55% NRMSE). This work demonstrates that it is possible to linearly speed up the inference of streaming CNNs processing overlapping windows, negating the additional computation typically incurred by overlapping windows.

Dimitrios Samakovlis, Stefano Albini, Rubén Rodríguez Álvarez et al.

The design of low-power wearables for the biomedical domain has received a lot of attention in recent decades, as technological advances in chip manufacturing have allowed real-time monitoring of patients using low-complexity ML within the mW range. Despite advances in application and hardware design research, the domain lacks a systematic approach to hardware evaluation. In this work, we propose BiomedBench, a new benchmark suite composed of complete end-to-end TinyML biomedical applications for real-time monitoring of patients using wearable devices. Each application presents different requirements during typical signal acquisition and processing phases, including varying computational workloads and relations between active and idle times. Furthermore, our evaluation of five state-of-the-art low-power platforms in terms of energy efficiency shows that modern platforms cannot effectively target all types of biomedical applications. BiomedBench is released as an open-source suite to standardize hardware evaluation and guide hardware and application design in the TinyML wearable domain.

Lara Orlandic, Jonathan Dan, Jerome Thevenot et al.

Chronic cough disorders are widespread and challenging to assess because they rely on subjective patient questionnaires about cough frequency. Wearable devices running Machine Learning (ML) algorithms are promising for quantifying daily coughs, providing clinicians with objective metrics to track symptoms and evaluate treatments. However, there is a mismatch between state-of-the-art metrics for cough counting algorithms and the information relevant to clinicians. Most works focus on distinguishing cough from non-cough samples, which does not directly provide clinically relevant outcomes such as the number of cough events or their temporal patterns. In addition, typical metrics such as specificity and accuracy can be biased by class imbalance. We propose using event-based evaluation metrics aligned with clinical guidelines on significant cough counting endpoints. We use an ML classifier to illustrate the shortcomings of traditional sample-based accuracy measurements, highlighting their variance due to dataset class imbalance and sample window length. We also present an open-source event-based evaluation framework to test algorithm performance in identifying cough events and rejecting false positives. We provide examples and best practice guidelines in event-based cough counting as a necessary first step to assess algorithm performance with clinical relevance.

Lara Orlandic, Tomas Teijeiro, David Atienza

Cough audio signal classification has been successfully used to diagnose a variety of respiratory conditions, and there has been significant interest in leveraging Machine Learning (ML) to provide widespread COVID-19 screening. However, there is currently no validated database of cough sounds with which to train such ML models. The COUGHVID dataset provides over 20,000 crowdsourced cough recordings representing a wide range of subject ages, genders, geographic locations, and COVID-19 statuses. First, we filtered the dataset using our open-sourced cough detection algorithm. Second, experienced pulmonologists labeled more than 2,000 recordings to diagnose medical abnormalities present in the coughs, thereby contributing one of the largest expert-labeled cough datasets in existence that can be used for a plethora of cough audio classification tasks. Finally, we ensured that coughs labeled as symptomatic and COVID-19 originate from countries with high infection rates, and that their expert labels are consistent. As a result, the COUGHVID dataset contributes a wealth of cough recordings for training ML models to address the world's most urgent health crises.

Karan Pathak, David Atienza, Marina Zapater

The growing demands in the training and inference of Large Language Models (LLMs) are accelerating the adoption of scale-up systems that extend server shared memory through the use of Compute Express Link (CXL)-based load/store interconnects. Accurate full-system simulation of such architectures remains challenging, as existing tools (all very recent) rely on simplified or non-compliant architectural models, impacting accuracy and usability. We present CXLRAMSim, the first gem5-integrated, full-system simulator that models CXL devices at their correct position on the I/O bus, enabling the use of unmodified Linux kernels and software stack, realistic latency-bandwidth behavior and true interleaving with system DRAM. Our approach provides high-fidelity CXL.mem characterization and captures key challenges such as cache pollution when accessing CXL memory.

Christodoulos Kechris, Jonathan Dan, Jose Miranda et al.

Accurate extraction of heart rate from photoplethysmography (PPG) signals remains challenging due to motion artifacts and signal degradation. Although deep learning methods trained as a data-driven inference problem offer promising solutions, they often underutilize existing knowledge from the medical and signal processing community. In this paper, we address three shortcomings of deep learning models: motion artifact removal, degradation assessment, and physiologically plausible analysis of the PPG signal. We propose KID-PPG, a knowledge-informed deep learning model that integrates expert knowledge through adaptive linear filtering, deep probabilistic inference, and data augmentation. We evaluate KID-PPG on the PPGDalia dataset, achieving an average mean absolute error of 2.85 beats per minute, surpassing existing reproducible methods. Our results demonstrate a significant performance improvement in heart rate tracking through the incorporation of prior knowledge into deep learning models. This approach shows promise in enhancing various biomedical applications by incorporating existing expert knowledge in deep learning models.

Hojjat Karami, Mary-Anne Hartley, David Atienza et al.

Time series in Electronic Health Records (EHRs) present unique challenges for generative models, such as irregular sampling, missing values, and high dimensionality. In this paper, we propose a novel generative adversarial network (GAN) model, TimEHR, to generate time series data from EHRs. In particular, TimEHR treats time series as images and is based on two conditional GANs. The first GAN generates missingness patterns, and the second GAN generates time series values based on the missingness pattern. Experimental results on three real-world EHR datasets show that TimEHR outperforms state-of-the-art methods in terms of fidelity, utility, and privacy metrics.

Hojjat Karami, David Atienza, Anisoara Ionescu

Irregular sampling of time series in electronic health records (EHRs) is one of the main challenges for developing machine learning models. Additionally, the pattern of missing data in certain clinical variables is not at random but depends on the decisions of clinicians and the state of the patient. Point process is a mathematical framework for analyzing event sequence data that is consistent with irregular sampling patterns. Our model, TEE4EHR, is a transformer event encoder (TEE) with point process loss that encodes the pattern of laboratory tests in EHRs. The utility of our TEE has been investigated in a variety of benchmark event sequence datasets. Additionally, we conduct experiments on two real-world EHR databases to provide a more comprehensive evaluation of our model. Firstly, in a self-supervised learning approach, the TEE is jointly learned with an existing attention-based deep neural network which gives superior performance in negative log-likelihood and future event prediction. Besides, we propose an algorithm for aggregating attention weights that can reveal the interaction between the events. Secondly, we transfer and freeze the learned TEE to the downstream task for the outcome prediction, where it outperforms state-of-the-art models for handling irregularly sampled time series. Furthermore, our results demonstrate that our approach can improve representation learning in EHRs and can be useful for clinical prediction tasks.

Hojjat Karami, David Atienza, Anisoara Ionescu

Generating synthetic Electronic Health Records (EHRs) offers significant potential for data augmentation, privacy-preserving data sharing, and improving machine learning model training. We propose a novel tokenization strategy tailored for structured EHR data, which encompasses diverse data types such as covariates, ICD codes, and irregularly sampled time series. Using a GPT-like decoder-only transformer model, we demonstrate the generation of high-quality synthetic EHRs. Our approach is evaluated using the MIMIC-III dataset, and we benchmark the fidelity, utility, and privacy of the generated data against state-of-the-art models.

Pedro Palacios, Rafael Medina, Jean-Luc Rouas et al.

Efficient deployment of resource-intensive transformers on edge devices necessitates cross-stack optimization. We thus study the interrelation between structured pruning and systolic acceleration, matching the size of pruned blocks with the systolic array dimensions. In this setting, computations of pruned weight blocks can be skipped, reducing run-time and energy consumption, but potentially impacting quality of service (QoS). To evaluate the trade-offs between systolic array size and sparsity opportunities, we present a novel co-design framework that integrates algorithmic optimization, system simulation, and hardware design. Targeting speech recognition and machine translation using transformers as case study, we analyze how configuration choices across the stack affect performance metrics. Results demonstrate that structured pruning on systems featuring systolic array acceleration can effectively increase performance, while maintaining high QoS levels. Up to 44% system-wide speedups due to structured pruning and quantization were measured, with only 1.4% word error rate degradation on the standard LibriSpeech dataset.

Jonathan Dan, Amirhossein Shahbazinia, Christodoulos Kechris et al.

Reliable automatic seizure detection from long-term EEG remains a challenge, as current machine learning models often fail to generalize across patients or clinical settings. Manual EEG review remains the clinical standard, underscoring the need for robust models and standardized evaluation. To rigorously assess algorithm performance, we organized a challenge using a private dataset of continuous EEG recordings from 65 subjects (4,360 hours). Expert neurophysiologists annotated the data, providing ground truth for seizure events. Participants were required to detect seizure onset and duration, with evaluation based on event-based metrics, including sensitivity, precision, F1-score, and false positives per day. The SzCORE framework ensured standardized evaluation. The primary ranking criterion was the event-based F1-score, reflecting clinical relevance by balancing sensitivity and false positives. The challenge received 30 submissions from 19 teams, with 28 algorithms evaluated. Results revealed wide variability in performance, with a top F1-score of 43% (sensitivity 37%, precision 45%), highlighting the ongoing difficulty of seizure detection. The challenge also revealed a gap between reported performance and real-world evaluation, emphasizing the importance of rigorous benchmarking. Compared to previous challenges and commercial systems, the best-performing algorithm in this contest showed improved performance. Importantly, the challenge platform now supports continuous benchmarking, enabling reproducible research, integration of new datasets, and clinical evaluation of seizure detection algorithms using a standardized framework.

Paula Ruiz-Barroso, Francisco M. Castro, José Miranda et al.

Cardiovascular diseases, a leading cause of noncommunicable disease-related deaths, require early and accurate detection to improve patient outcomes. Taking advantage of advances in machine learning and deep learning, multiple approaches have been proposed in the literature to address the challenge of detecting ECG anomalies. Typically, these methods are based on the manual interpretation of ECG signals, which is time consuming and depends on the expertise of healthcare professionals. The objective of this work is to propose a deep learning system, FADE, designed for normal ECG forecasting and anomaly detection, which reduces the need for extensive labeled datasets and manual interpretation. FADE has been trained in a self-supervised manner with a novel morphological inspired loss function. Unlike conventional models that learn from labeled anomalous ECG waveforms, our approach predicts the future of normal ECG signals, thus avoiding the need for extensive labeled datasets. Using a novel distance function to compare forecasted ECG signals with actual sensor data, our method effectively identifies cardiac anomalies. Additionally, this approach can be adapted to new contexts through domain adaptation techniques. To evaluate our proposal, we performed a set of experiments using two publicly available datasets: MIT-BIH NSR and MIT-BIH Arrythmia. The results demonstrate that our system achieves an average accuracy of 83.84% in anomaly detection, while correctly classifying normal ECG signals with an accuracy of 85.46%. Our proposed approach exhibited superior performance in the early detection of cardiac anomalies in ECG signals, surpassing previous methods that predominantly identify a limited range of anomalies. FADE effectively detects both abnormal heartbeats and arrhythmias, offering significant advantages in healthcare through cost reduction or processing of large-scale ECG data.

Alireza Amirshahi, Giovanni Ansaloni, David Atienza

The increasing complexity of transformer models in artificial intelligence expands their computational costs, memory usage, and energy consumption. Hardware acceleration tackles the ensuing challenges by designing processors and accelerators tailored for transformer models, supporting their computation hotspots with high efficiency. However, memory bandwidth can hinder improvements in hardware accelerators. Against this backdrop, in this paper we propose a novel memory arrangement strategy, governed by the hardware accelerator's kernel size, which effectively minimizes off-chip data access. This arrangement is particularly beneficial for end-to-end transformer model inference, where most of the computation is based on general matrix multiplication (GEMM) operations. Additionally, we address the overhead of non-GEMM operations in transformer models within the scope of this memory data arrangement. Our study explores the implementation and effectiveness of the proposed accelerator-driven data arrangement approach in both single- and multi-core systems. Our evaluation demonstrates that our approach can achieve up to a 2.8x speed increase when executing inferences employing state-of-the-art transformers.

Deniz Kasap, Taraneh Aminosharieh Najafi, Jérôme Paul Rémy Thevenot et al.

We present VersaPants, the first loose-fitting, textile-based capacitive sensing system for lower-body motion capture, built on the open-hardware VersaSens platform. By integrating conductive textile patches and a compact acquisition unit into a pair of pants, the system reconstructs lower-body pose without compromising comfort. Unlike IMU-based systems that require user-specific fitting or camera-based methods that compromise privacy, our approach operates without fitting adjustments and preserves user privacy. VersaPants is a custom-designed smart garment featuring 6 capacitive channels per leg. We employ a lightweight Transformer-based deep learning model that maps capacitance signals to joint angles, enabling embedded implementation on edge platforms. To test our system, we collected approximately 3.7 hours of motion data from 11 participants performing 16 daily and exercise-based movements. The model achieves a mean per-joint position error (MPJPE) of 11.96 cm and a mean per-joint angle error (MPJAE) of 12.3 degrees across the hip, knee, and ankle joints, indicating the model's ability to generalize to unseen users and movements. A comparative analysis of existing textile-based deep learning architectures reveals that our model achieves competitive reconstruction performance with up to 22 times fewer parameters and 18 times fewer FLOPs, enabling real-time inference at 42 FPS on a commercial smartwatch without quantization. These results position VersaPants as a promising step toward scalable, comfortable, and embedded motion-capture solutions for fitness, healthcare, and wellbeing applications.

Amirhossein Shahbazinia, Jonathan Dan, Jose A. Miranda et al.

Objective: Epilepsy, a prevalent neurological disease, demands careful diagnosis and continuous care. Seizure detection remains challenging, as current clinical practice relies on expert analysis of electroencephalography, which is a time-consuming process and requires specialized knowledge. Addressing this challenge, this paper explores automated epileptic seizure detection using deep learning, focusing on personalized continual learning models that adapt to each patient's unique electroencephalography signal features, which evolve over time. Methods: In this context, our approach addresses the challenge of integrating new data into existing models without catastrophic forgetting, a common issue in static deep learning models. We propose EpiSMART, a continual learning framework for seizure detection that uses a size-constrained replay buffer and an informed sample selection strategy to incrementally adapt to patient-specific electroencephalography signals. By selectively retaining high-entropy and seizure-predicted samples, our method preserves critical past information while maintaining high performance with minimal memory and computational requirements. Results: Validation on the CHB-MIT dataset, shows that EpiSMART achieves a 21% improvement in the F1 score over a trained baseline without updates in all other patients. On average, EpiSMART requires only 6.46 minutes of labeled data and 6.28 updates per day, making it suitable for real-time deployment in wearable systems. Conclusion:EpiSMART enables robust and personalized seizure detection under realistic and resource-constrained conditions by effectively integrating new data into existing models without degrading past knowledge. Significance: This framework advances automated seizure detection by providing a continual learning approach that supports patient-specific adaptation and practical deployment in wearable healthcare systems.

Golnoosh Abdollahinejad, Saleh Baghersalimi, Denisa-Andreea Constantinescu et al.

This work addresses the challenge of forecasting urban water dynamics by developing a multi-input, multi-output deep learning model that incorporates both endogenous variables (e.g., water height or discharge) and exogenous factors (e.g., precipitation history and forecast reports). Unlike conventional forecasting, the proposed model, AquaCast, captures both inter-variable and temporal dependencies across all inputs, while focusing forecast solely on endogenous variables. Exogenous inputs are fused via an embedding layer, eliminating the need to forecast them and enabling the model to attend to their short-term influences more effectively. We evaluate our approach on the LausanneCity dataset, which includes measurements from four urban drainage sensors, and demonstrate state-of-the-art performance when using only endogenous variables. Performance also improves with the inclusion of exogenous variables and forecast reports. To assess generalization and scalability, we additionally test the model on three large-scale synthesized datasets, generated from MeteoSwiss records, the Lorenz Attractors model, and the Random Fields model, each representing a different level of temporal complexity across 100 nodes. The results confirm that our model consistently outperforms existing baselines and maintains a robust and accurate forecast across both real and synthetic datasets.

Amirhossein Shahbazinia, Darong Huang, Luis Costero et al.

Cloud platforms are increasingly relied upon to host diverse, resource-intensive workloads due to their scalability, flexibility, and cost-efficiency. In multi-tenant cloud environments, virtual machines are consolidated on shared physical servers to improve resource utilization. While virtualization guarantees resource partitioning for CPU, memory, and storage, it cannot ensure performance isolation. Competition for shared resources such as last-level cache, memory bandwidth, and network interfaces often leads to severe performance degradation. Existing management techniques, including VM scheduling and resource provisioning, require accurate performance prediction to mitigate interference. However, this remains challenging in public clouds due to the black-box nature of VMs and the highly dynamic nature of workloads. To address these limitations, we propose CloudFormer, a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. CloudFormer jointly models temporal dynamics and system-level interactions, leveraging 206 system metrics at one-second resolution across both static and dynamic scenarios. This design enables the model to capture transient interference effects and adapt to varying workload conditions without scenario-specific tuning. Complementing the methodology, we provide a fine-grained dataset that significantly expands the temporal resolution and metric diversity compared to existing benchmarks. Experimental results demonstrate that CloudFormer consistently outperforms state-of-the-art baselines across multiple evaluation metrics, achieving robust generalization across diverse and previously unseen workloads. Notably, CloudFormer attains a mean absolute error (MAE) of just 7.8%, representing a substantial improvement in predictive accuracy and outperforming existing methods at least by 28%.

Giulio Masinelli, Chang Rajani, Patrik Hoffmann et al.

Ensuring consistent processing quality is challenging in laser processes due to varying material properties and surface conditions. Although some approaches have shown promise in solving this problem via automation, they often rely on predetermined targets or are limited to simulated environments. To address these shortcomings, we propose a novel real-time reinforcement learning approach for laser process control, implemented on a Field Programmable Gate Array to achieve real-time execution. Our experimental results from laser welding tests on stainless steel samples with a range of surface roughnesses validated the method's ability to adapt autonomously, without relying on reward engineering or prior setup information. Specifically, the algorithm learned the correct power profile for each unique surface characteristic, demonstrating significant improvements over hand-engineered optimal constant power strategies -- up to 23% better performance on rougher surfaces and 7% on mixed surfaces. This approach represents a significant advancement in automating and optimizing laser processes, with potential applications across multiple industries.

Una Pale, Tomas Teijeiro, David Atienza

Wearable and unobtrusive monitoring and prediction of epileptic seizures has the potential to significantly increase the life quality of patients, but is still an unreached goal due to challenges of real-time detection and wearable devices design. Hyperdimensional (HD) computing has evolved in recent years as a new promising machine learning approach, especially when talking about wearable applications. But in the case of epilepsy detection, standard HD computing is not performing at the level of other state-of-the-art algorithms. This could be due to the inherent complexity of the seizures and their signatures in different biosignals, such as the electroencephalogram (EEG), the highly personalized nature, and the disbalance of seizure and non-seizure instances. In the literature, different strategies for improved learning of HD computing have been proposed, such as iterative (multi-pass) learning, multi-centroid learning and learning with sample weight ("OnlineHD"). Yet, most of them have not been tested on the challenging task of epileptic seizure detection, and it stays unclear whether they can increase the HD computing performance to the level of the current state-of-the-art algorithms, such as random forests. Thus, in this paper, we implement different learning strategies and assess their performance on an individual basis, or in combination, regarding detection performance and memory and computational requirements. Results show that the best-performing algorithm, which is a combination of multi-centroid and multi-pass, can indeed reach the performance of the random forest model on a highly unbalanced dataset imitating a real-life epileptic seizure detection application.

Elisabetta De Giovanni, Tomas Teijeiro, Grégoire P. Millet et al.

Objective: Continuous monitoring of biosignals via wearable sensors has quickly expanded in the medical and wellness fields. At rest, automatic detection of vital parameters is generally accurate. However, in conditions such as high-intensity exercise, sudden physiological changes occur to the signals, compromising the robustness of standard algorithms. Methods: Our method, called BayeSlope, is based on unsupervised learning, Bayesian filtering, and non-linear normalization to enhance and correctly detect the R peaks according to their expected positions in the ECG. Furthermore, as BayeSlope is computationally heavy and can drain the device battery quickly, we propose an online design that adapts its robustness to sudden physiological changes, and its complexity to the heterogeneous resources of modern embedded platforms. This method combines BayeSlope with a lightweight algorithm, executed in cores with different capabilities, to reduce the energy consumption while preserving the accuracy. Results: BayeSlope achieves an F1 score of 99.3% in experiments during intense cycling exercise with 20 subjects. Additionally, the online adaptive process achieves an F1 score of 99% across five different exercise intensities, with a total energy consumption of 1.55+-0.54~mJ. Conclusion: We propose a highly accurate and robust method, and a complete energy-efficient implementation in a modern ultra-low-power embedded platform to improve R peak detection in challenging conditions, such as during high-intensity exercise. Significance: The experiments show that BayeSlope outperforms a state-of-the-art algorithm up to 8.4% in F1 score, while our online adaptive method can reach energy savings up to 38.7% on modern heterogeneous wearable platforms.

Una Pale, Tomas Teijeiro, David Atienza

Long-term monitoring of patients with epilepsy presents a challenging problem from the engineering perspective of real-time detection and wearable devices design. It requires new solutions that allow continuous unobstructed monitoring and reliable detection and prediction of seizures. A high variability in the electroencephalogram (EEG) patterns exists among people, brain states, and time instances during seizures, but also during non-seizure periods. This makes epileptic seizure detection very challenging, especially if data is grouped under only seizure and non-seizure labels. Hyperdimensional (HD) computing, a novel machine learning approach, comes in as a promising tool. However, it has certain limitations when the data shows a high intra-class variability. Therefore, in this work, we propose a novel semi-supervised learning approach based on a multi-centroid HD computing. The multi-centroid approach allows to have several prototype vectors representing seizure and non-seizure states, which leads to significantly improved performance when compared to a simple 2-class HD model. Further, real-life data imbalance poses an additional challenge and the performance reported on balanced subsets of data is likely to be overestimated. Thus, we test our multi-centroid approach with three different dataset balancing scenarios, showing that performance improvement is higher for the less balanced dataset. More specifically, up to 14% improvement is achieved on an unbalanced test set with 10 times more non-seizure than seizure data. At the same time, the total number of sub-classes is not significantly increased compared to the balanced dataset. Thus, the proposed multi-centroid approach can be an important element in achieving a high performance of epilepsy detection with real-life data balance or during online learning, where seizures are infrequent.

David Atienza, Concha Bielza, Pedro Larrañaga

We introduce semiparametric Bayesian networks that combine parametric and nonparametric conditional probability distributions. Their aim is to incorporate the advantages of both components: the bounded complexity of parametric models and the flexibility of nonparametric ones. We demonstrate that semiparametric Bayesian networks generalize two well-known types of Bayesian networks: Gaussian Bayesian networks and kernel density estimation Bayesian networks. For this purpose, we consider two different conditional probability distributions required in a semiparametric Bayesian network. In addition, we present modifications of two well-known algorithms (greedy hill-climbing and PC) to learn the structure of a semiparametric Bayesian network from data. To realize this, we employ a score function based on cross-validation. In addition, using a validation dataset, we apply an early-stopping criterion to avoid overfitting. To evaluate the applicability of the proposed algorithm, we conduct an exhaustive experiment on synthetic data sampled by mixing linear and nonlinear functions, multivariate normal data sampled from Gaussian Bayesian networks, real data from the UCI repository, and bearings degradation data. As a result of this experiment, we conclude that the proposed algorithm accurately learns the combination of parametric and nonparametric components, while achieving a performance comparable with those provided by state-of-the-art methods.

Una Pale, Tomas Teijeiro, David Atienza

Hyperdimensional computing is a promising novel paradigm for low-power embedded machine learning. It has been applied on different biomedical applications, and particularly on epileptic seizure detection. Unfortunately, due to differences in data preparation, segmentation, encoding strategies, and performance metrics, results are hard to compare, which makes building upon that knowledge difficult. Thus, the main goal of this work is to perform a systematic assessment of the HD computing framework for the detection of epileptic seizures, comparing different feature approaches mapped to HD vectors. More precisely, we test two previously implemented features as well as several novel approaches with HD computing on epileptic seizure detection. We evaluate them in a comparable way, i.e., with the same preprocessing setup, and with the identical performance measures. We use two different datasets in order to assess the generalizability of our conclusions. The systematic assessment involved three primary aspects relevant for potential wearable implementations: 1) detection performance, 2) memory requirements, and 3) computational complexity. Our analysis shows a significant difference in detection performance between approaches, but also that the ones with the highest performance might not be ideal for wearable applications due to their high memory or computational requirements. Furthermore, we evaluate a post-processing strategy to adjust the predictions to the dynamics of epileptic seizures, showing that performance is significantly improved in all the approaches and also that after post-processing, differences in performance are much smaller between approaches.

Lara Orlandic, Adriana Arza Valdes, David Atienza

A person's passage of time perception (POTP) is strongly linked to their mental state and stress response, and can therefore provide an easily quantifiable means of continuous mental health monitoring. In this work, we develop a custom experiment and Machine Learning (ML) models for predicting POTP from biomarkers acquired from wearable biosensors. We first confirm that individuals experience time passing slower than usual during fear or sadness (p = 0.046) and faster than usual during cognitive tasks (p = 2 x 10^-5). Then, we group together the experimental segments associated with fast, slow, and normal POTP, and train a ML model to classify between these states based on a person's biomarkers. The classifier had a weighted average F-1 score of 79%, with the fast-passing time class having the highest F-1 score of 93%. Next, we classify each individual's POTP regardless of the task at hand, achieving an F-1 score of 77.1% when distinguishing time passing faster rather than slower than usual. In the two classifiers, biomarkers derived from the respiration, electrocardiogram, skin conductance, and skin temperature signals contributed most to the classifier output, thus enabling real-time POTP monitoring using noninvasive, wearable biosensors.

Arman Iranfar, Adriana Arza, David Atienza

Continuous and multimodal stress detection has been performed recently through wearable devices and machine learning algorithms. However, a well-known and important challenge of working on physiological signals recorded by conventional monitoring devices is missing data due to sensors insufficient contact and interference by other equipment. This challenge becomes more problematic when the user/patient is mentally or physically active or stressed because of more frequent conscious or subconscious movements. In this paper, we propose ReLearn, a robust machine learning framework for stress detection from biomarkers extracted from multimodal physiological signals. ReLearn effectively copes with missing data and outliers both at training and inference phases. ReLearn, composed of machine learning models for feature selection, outlier detection, data imputation, and classification, allows us to classify all samples, including those with missing values at inference. In particular, according to our experiments and stress database, while by discarding all missing data, as a simplistic yet common approach, no prediction can be made for 34% of the data at inference, our approach can achieve accurate predictions, as high as 78%, for missing samples. Also, our experiments show that the proposed framework obtains a cross-validation accuracy of 86.8% even if more than 50% of samples within the features are missing.

Valentin Gabeff, Tomas Teijeiro, Marina Zapater et al.

While Deep Learning (DL) is often considered the state-of-the art for Artificial Intelligence-based medical decision support, it remains sparsely implemented in clinical practice and poorly trusted by clinicians due to insufficient interpretability of neural network models. We have tackled this issue by developing interpretable DL models in the context of online detection of epileptic seizure, based on EEG signal. This has conditioned the preparation of the input signals, the network architecture, and the post-processing of the output in line with the domain knowledge. Specifically, we focused the discussion on three main aspects: 1) how to aggregate the classification results on signal segments provided by the DL model into a larger time scale, at the seizure-level; 2) what are the relevant frequency patterns learned in the first convolutional layer of different models, and their relation with the delta, theta, alpha, beta and gamma frequency bands on which the visual interpretation of EEG is based; and 3) the identification of the signal waveforms with larger contribution towards the ictal class, according to the activation differences highlighted using the DeepLIFT method. Results show that the kernel size in the first layer determines the interpretability of the extracted features and the sensitivity of the trained models, even though the final performance is very similar after post-processing. Also, we found that amplitude is the main feature leading to an ictal prediction, suggesting that a larger patient population would be required to learn more complex frequency patterns. Still, our methodology was successfully able to generalize patient inter-variability for the majority of the studied population with a classification F1-score of 0.873 and detecting 90% of the seizures.

Damian Pascual, Amir Aminifar, David Atienza et al.

Epilepsy is a chronic neurological disorder affecting more than 65 million people worldwide and manifested by recurrent unprovoked seizures. The unpredictability of seizures not only degrades the quality of life of the patients, but it can also be life-threatening. Modern systems monitoring electroencephalography (EEG) signals are being currently developed with the view to detect epileptic seizures in order to alert caregivers and reduce the impact of seizures on patients' quality of life. Such seizure detection systems employ state-of-the-art machine learning algorithms that require a considerably large amount of labeled personal data for training. However, acquiring EEG signals of epileptic seizures is a costly and time-consuming process for medical experts and patients, currently requiring in-hospital recordings in specialized units. In this work, we generate synthetic seizure-like brain electrical activities, i.e., EEG signals, that can be used to train seizure detection algorithms, alleviating the need for recorded data. First, we train a Generative Adversarial Network (GAN) with data from 30 epilepsy patients. Then, we generate synthetic personalized training sets for new, unseen patients, which overall yield higher detection performance than the real-data training sets. We demonstrate our results using the datasets from the EPILEPSIAE Project, one of the world's largest public databases for seizure detection.