Cecilia Mascolo

Dimitris Spathis, Ignacio Perez-Pozuelo, Tomas I. Gonzales et al. · cambridge

Cardiorespiratory fitness is an established predictor of metabolic disease and mortality. Fitness is directly measured as maximal oxygen consumption (VO$_{2}max$), or indirectly assessed using heart rate responses to standard exercise tests. However, such testing is costly and burdensome because it requires specialized equipment such as treadmills and oxygen masks, limiting its utility. Modern wearables capture dynamic real-world data which could improve fitness prediction. In this work, we design algorithms and models that convert raw wearable sensor data into cardiorespiratory fitness estimates. We validate these estimates' ability to capture fitness profiles in free-living conditions using the Fenland Study (N=11,059), along with its longitudinal cohort (N=2,675), and a third external cohort using the UK Biobank Validation Study (N=181) who underwent maximal VO$_{2}max$ testing, the gold standard measurement of fitness. Our results show that the combination of wearables and other biomarkers as inputs to neural networks yields a strong correlation to ground truth in a holdout sample (r = 0.82, 95CI 0.80-0.83), outperforming other approaches and models and detects fitness change over time (e.g., after 7 years). We also show how the model's latent space can be used for fitness-aware patient subtyping paving the way to scalable interventions and personalized trial recruitment. These results demonstrate the value of wearables for fitness estimation that today can be measured only with laboratory tests.

Yu Wu, Dimitris Spathis, Hong Jia et al. · cambridge

Deep learning models have shown great promise in various healthcare monitoring applications. However, most healthcare datasets with high-quality (gold-standard) labels are small-scale, as directly collecting ground truth is often costly and time-consuming. As a result, models developed and validated on small-scale datasets often suffer from overfitting and do not generalize well to unseen scenarios. At the same time, large amounts of imprecise (silver-standard) labeled data, annotated by approximate methods with the help of modern wearables and in the absence of ground truth validation, are starting to emerge. However, due to measurement differences, this data displays significant label distribution shifts, which motivates the use of domain adaptation. To this end, we introduce UDAMA, a method with two key components: Unsupervised Domain Adaptation and Multidiscriminator Adversarial Training, where we pre-train on the silver-standard data and employ adversarial adaptation with the gold-standard data along with two domain discriminators. In particular, we showcase the practical potential of UDAMA by applying it to Cardio-respiratory fitness (CRF) prediction. CRF is a crucial determinant of metabolic disease and mortality, and it presents labels with various levels of noise (goldand silver-standard), making it challenging to establish an accurate prediction model. Our results show promising performance by alleviating distribution shifts in various label shift settings. Additionally, by using data from two free-living cohort studies (Fenland and BBVS), we show that UDAMA consistently outperforms up to 12% compared to competitive transfer learning and state-of-the-art domain adaptation models, paving the way for leveraging noisy labeled data to improve fitness estimation at scale.

Yu Wu, Dimitris Spathis, Hong Jia et al. · cambridge

Deep learning models have shown great promise in various healthcare applications. However, most models are developed and validated on small-scale datasets, as collecting high-quality (gold-standard) labels for health applications is often costly and time-consuming. As a result, these models may suffer from overfitting and not generalize well to unseen data. At the same time, an extensive amount of data with imprecise labels (silver-standard) is starting to be generally available, as collected from inexpensive wearables like accelerometers and electrocardiography sensors. These currently underutilized datasets and labels can be leveraged to produce more accurate clinical models. In this work, we propose UDAMA, a novel model with two key components: Unsupervised Domain Adaptation and Multi-discriminator Adversarial training, which leverage noisy data from source domain (the silver-standard dataset) to improve gold-standard modeling. We validate our framework on the challenging task of predicting lab-measured maximal oxygen consumption (VO$_{2}$max), the benchmark metric of cardio-respiratory fitness, using free-living wearable sensor data from two cohort studies as inputs. Our experiments show that the proposed framework achieves the best performance of corr = 0.665 $\pm$ 0.04, paving the way for accurate fitness estimation at scale.

Young D. Kwon, Jagmohan Chauhan, Hong Jia et al.

Continual Learning (CL) allows applications such as user personalization and household robots to learn on the fly and adapt to context. This is an important feature when context, actions, and users change. However, enabling CL on resource-constrained embedded systems is challenging due to the limited labeled data, memory, and computing capacity. In this paper, we propose LifeLearner, a hardware-aware meta continual learning system that drastically optimizes system resources (lower memory, latency, energy consumption) while ensuring high accuracy. Specifically, we (1) exploit meta-learning and rehearsal strategies to explicitly cope with data scarcity issues and ensure high accuracy, (2) effectively combine lossless and lossy compression to significantly reduce the resource requirements of CL and rehearsal samples, and (3) developed hardware-aware system on embedded and IoT platforms considering the hardware characteristics. As a result, LifeLearner achieves near-optimal CL performance, falling short by only 2.8% on accuracy compared to an Oracle baseline. With respect to the state-of-the-art (SOTA) Meta CL method, LifeLearner drastically reduces the memory footprint (by 178.7x), end-to-end latency by 80.8-94.2%, and energy consumption by 80.9-94.2%. In addition, we successfully deployed LifeLearner on two edge devices and a microcontroller unit, thereby enabling efficient CL on resource-constrained platforms where it would be impractical to run SOTA methods and the far-reaching deployment of adaptable CL in a ubiquitous manner. Code is available at https://github.com/theyoungkwon/LifeLearner.

Chi Ian Tang, Lorena Qendro, Dimitris Spathis et al. · cambridge

Self-supervised learning (SSL) has shown remarkable performance in computer vision tasks when trained offline. However, in a Continual Learning (CL) scenario where new data is introduced progressively, models still suffer from catastrophic forgetting. Retraining a model from scratch to adapt to newly generated data is time-consuming and inefficient. Previous approaches suggested re-purposing self-supervised objectives with knowledge distillation to mitigate forgetting across tasks, assuming that labels from all tasks are available during fine-tuning. In this paper, we generalize self-supervised continual learning in a practical setting where available labels can be leveraged in any step of the SSL process. With an increasing number of continual tasks, this offers more flexibility in the pre-training and fine-tuning phases. With Kaizen, we introduce a training architecture that is able to mitigate catastrophic forgetting for both the feature extractor and classifier with a carefully designed loss function. By using a set of comprehensive evaluation metrics reflecting different aspects of continual learning, we demonstrated that Kaizen significantly outperforms previous SSL models in competitive vision benchmarks, with up to 16.5% accuracy improvement on split CIFAR-100. Kaizen is able to balance the trade-off between knowledge retention and learning from new data with an end-to-end model, paving the way for practical deployment of continual learning systems.

Sotirios Vavaroutas, Yu Yvonne Wu, Ali Etemad et al.

Data samples used for training often differ from those encountered during fine-tuning and deployment, and while ML models show promise, their performance remains limited when only small annotated datasets are available. Performance often degrades under distribution shifts caused by diverse sensors, populations, and application settings. Although pre-training helps, models frequently encounter out-of-distribution (OOD) data in real-world settings, leading to reduced robustness. Existing adaptation methods usually assume fixed distribution shifts and struggle when multiple types or severities occur. In particular, they overlook shift severity, for example treating adaptation to a large familiar dataset the same as adaptation to a small dataset with a new task, which limits generalisation. To address this, we propose ADAPTOOD, a novel framework that leverages data uncertainty to quantify distribution shift severity and guide fine-tuning for time series. This uncertainty measures how strongly samples from the target deployment distribution deviate from the pre-training distribution, providing a direct signal of OOD severity. Our framework combines this uncertainty with low-rank model updates and adaptive hyperparameter optimisation to improve adaptation. We show that ADAPTOOD achieves up to 7% higher accuracy and 12.9% higher precision than existing methods in OOD tasks, maintaining strong performance as distribution shift severity increases.

Young D. Kwon, Rui Li, Stylianos I. Venieris et al.

On-device training is essential for user personalisation and privacy. With the pervasiveness of IoT devices and microcontroller units (MCUs), this task becomes more challenging due to the constrained memory and compute resources, and the limited availability of labelled user data. Nonetheless, prior works neglect the data scarcity issue, require excessively long training time (e.g. a few hours), or induce substantial accuracy loss (>10%). In this paper, we propose TinyTrain, an on-device training approach that drastically reduces training time by selectively updating parts of the model and explicitly coping with data scarcity. TinyTrain introduces a task-adaptive sparse-update method that dynamically selects the layer/channel to update based on a multi-objective criterion that jointly captures user data, the memory, and the compute capabilities of the target device, leading to high accuracy on unseen tasks with reduced computation and memory footprint. TinyTrain outperforms vanilla fine-tuning of the entire network by 3.6-5.0% in accuracy, while reducing the backward-pass memory and computation cost by up to 1,098x and 7.68x, respectively. Targeting broadly used real-world edge devices, TinyTrain achieves 9.5x faster and 3.5x more energy-efficient training over status-quo approaches, and 2.23x smaller memory footprint than SOTA methods, while remaining within the 1 MB memory envelope of MCU-grade platforms.

Young D. Kwon, Jagmohan Chauhan, Cecilia Mascolo

With the advancement of Deep Neural Networks (DNN) and large amounts of sensor data from Internet of Things (IoT) systems, the research community has worked to reduce the computational and resource demands of DNN to compute on low-resourced microcontrollers (MCUs). However, most of the current work in embedded deep learning focuses on solving a single task efficiently, while the multi-tasking nature and applications of IoT devices demand systems that can handle a diverse range of tasks (activity, voice, and context recognition) with input from a variety of sensors, simultaneously. In this paper, we propose YONO, a product quantization (PQ) based approach that compresses multiple heterogeneous models and enables in-memory model execution and switching for dissimilar multi-task learning on MCUs. We first adopt PQ to learn codebooks that store weights of different models. Also, we propose a novel network optimization and heuristics to maximize the compression rate and minimize the accuracy loss. Then, we develop an online component of YONO for efficient model execution and switching between multiple tasks on an MCU at run time without relying on an external storage device. YONO shows remarkable performance as it can compress multiple heterogeneous models with negligible or no loss of accuracy up to 12.37$\times$. Besides, YONO's online component enables an efficient execution (latency of 16-159 ms per operation) and reduces model loading/switching latency and energy consumption by 93.3-94.5% and 93.9-95.0%, respectively, compared to external storage access. Interestingly, YONO can compress various architectures trained with datasets that were not shown during YONO's offline codebook learning phase showing the generalizability of our method. To summarize, YONO shows great potential and opens further doors to enable multi-task learning systems on extremely resource-constrained devices.

Feixiang Zheng, Yu Wu, Cecilia Mascolo et al.

Time series data from the Intensive Care Unit (ICU) provides critical information for patient monitoring. While recent advancements in applying Large Language Models (LLMs) to time series modeling (TSM) have shown great promise, their effectiveness on the irregular ICU data, characterized by particularly high rates of missing values, remains largely unexplored. This work investigates two key components underlying the success of LLMs for TSM: the time series encoder and the multimodal alignment strategy. To this end, we establish a systematic testbed to evaluate their impact across various state-of-the-art LLM-based methods on benchmark ICU datasets against strong supervised and self-supervised baselines. Results reveal that the encoder design is more critical than the alignment strategy. Encoders that explicitly model irregularity achieve substantial performance gains, yielding an average AUPRC increase of $12.8\%$ over the vanilla Transformer. While less impactful, the alignment strategy is also noteworthy, with the best-performing semantically rich, fusion-based strategy achieving a modest $2.9\%$ improvement over cross-attention. However, LLM-based methods require at least 10$\times$ longer training than the best-performing irregular supervised models, while delivering only comparable performance. They also underperform in data-scarce few-shot learning settings. These findings highlight both the promise and current limitations of LLMs for irregular ICU time series. The code is available at https://github.com/mHealthUnimelb/LLMTS.

Jialu Tang, Tong Xia, Yuan Lu et al.

Interpreting electrocardiograms (ECGs) and generating comprehensive reports remain challenging tasks in cardiology, often requiring specialized expertise and significant time investment. To address these critical issues, we propose ECG-ReGen, a retrieval-based approach for ECG-to-text report generation and question answering. Our method leverages a self-supervised learning for the ECG encoder, enabling efficient similarity searches and report retrieval. By combining pre-training with dynamic retrieval and Large Language Model (LLM)-based refinement, ECG-ReGen effectively analyzes ECG data and answers related queries, with the potential of improving patient care. Experiments conducted on the PTB-XL and MIMIC-IV-ECG datasets demonstrate superior performance in both in-domain and cross-domain scenarios for report generation. Furthermore, our approach exhibits competitive performance on ECG-QA dataset compared to fully supervised methods when utilizing off-the-shelf LLMs for zero-shot question answering. This approach, effectively combining self-supervised encoder and LLMs, offers a scalable and efficient solution for accurate ECG interpretation, holding significant potential to enhance clinical decision-making.

Tong Xia, Jing Han, Abhirup Ghosh et al.

Federated learning (FL) aided health diagnostic models can incorporate data from a large number of personal edge devices (e.g., mobile phones) while keeping the data local to the originating devices, largely ensuring privacy. However, such a cross-device FL approach for health diagnostics still imposes many challenges due to both local data imbalance (as extreme as local data consists of a single disease class) and global data imbalance (the disease prevalence is generally low in a population). Since the federated server has no access to data distribution information, it is not trivial to solve the imbalance issue towards an unbiased model. In this paper, we propose FedLoss, a novel cross-device FL framework for health diagnostics. Here the federated server averages the models trained on edge devices according to the predictive loss on the local data, rather than using only the number of samples as weights. As the predictive loss better quantifies the data distribution at a device, FedLoss alleviates the impact of data imbalance. Through a real-world dataset on respiratory sound and symptom-based COVID-$19$ detection task, we validate the superiority of FedLoss. It achieves competitive COVID-$19$ detection performance compared to a centralised model with an AUC-ROC of $79\%$. It also outperforms the state-of-the-art FL baselines in sensitivity and convergence speed. Our work not only demonstrates the promise of federated COVID-$19$ detection but also paves the way to a plethora of mobile health model development in a privacy-preserving fashion.

Dong Ma, Chi Ian Tang, Cecilia Mascolo

Many deep learning applications, like keyword spotting, require the incorporation of new concepts (classes) over time, referred to as Class Incremental Learning (CIL). The major challenge in CIL is catastrophic forgetting, i.e., preserving as much of the old knowledge as possible while learning new tasks. Various techniques, such as regularization, knowledge distillation, and the use of exemplars, have been proposed to resolve this issue. However, prior works primarily focus on the incremental learning step, while ignoring the optimization during the base model training. We hypothesize that a more transferable and generalizable feature representation from the base model would be beneficial to incremental learning. In this work, we adopt multitask learning during base model training to improve the feature generalizability. Specifically, instead of training a single model with all the base classes, we decompose the base classes into multiple subsets and regard each of them as a task. These tasks are trained concurrently and a shared feature extractor is obtained for incremental learning. We evaluate our approach on two datasets under various configurations. The results show that our approach enhances the average incremental learning accuracy by up to 5.5%, which enables more reliable and accurate keyword spotting over time. Moreover, the proposed approach can be combined with many existing techniques and provides additional performance gain.

Yu Yvonne Wu, Yuwei Zhang, Hyungjun Yoon et al.

Wearable foundation models (WFMs), trained on large volumes of data collected by affordable, always-on devices, have demonstrated strong performance on short-term, well-defined health monitoring tasks, including activity recognition, fitness tracking, and cardiovascular signal assessment. However, most existing WFMs primarily map short temporal windows to predefined labels via static encoders, emphasizing retrospective prediction rather than reasoning over evolving personal history, context, and future risk trajectories. As a result, they are poorly suited for modeling chronic, progressive, or episodic health conditions that unfold over weeks, months or years. Hence, we argue that WFMs must move beyond static encoders and be explicitly designed for longitudinal, anticipatory health reasoning. We identify three foundational shifts required to enable this transition: (1) Structurally rich data, which goes beyond isolated datasets or outcome-conditioned collection to integrated multimodal, long-term personal trajectories, and contextual metadata, ideally supported by open and interoperable data ecosystems; (2) Longitudinal-aware multimodal modeling, which prioritizes long-context inference, temporal abstraction, and personalization over cross-sectional or population-level prediction; and (3) Agentic inference systems, which move beyond static prediction to support planning, decision-making, and clinically grounded intervention under uncertainty. Together, these shifts reframe wearable health monitoring from retrospective signal interpretation toward continuous, anticipatory, and human-aligned health support.

Tong Xia, Abhirup Ghosh, Xinchi Qiu et al.

Federated Learning (FL) enables model development by leveraging data distributed across numerous edge devices without transferring local data to a central server. However, existing FL methods still face challenges when dealing with scarce and label-skewed data across devices, resulting in local model overfitting and drift, consequently hindering the performance of the global model. In response to these challenges, we propose a pioneering framework called \textit{FLea}, incorporating the following key components: \textit{i)} A global feature buffer that stores activation-target pairs shared from multiple clients to support local training. This design mitigates local model drift caused by the absence of certain classes; \textit{ii)} A feature augmentation approach based on local and global activation mix-ups for local training. This strategy enlarges the training samples, thereby reducing the risk of local overfitting; \textit{iii)} An obfuscation method to minimize the correlation between intermediate activations and the source data, enhancing the privacy of shared features. To verify the superiority of \textit{FLea}, we conduct extensive experiments using a wide range of data modalities, simulating different levels of local data scarcity and label skew. The results demonstrate that \textit{FLea} consistently outperforms state-of-the-art FL counterparts (among 13 of the experimented 18 settings, the improvement is over $5\%$) while concurrently mitigating the privacy vulnerabilities associated with shared features. Code is available at https://github.com/XTxiatong/FLea.git

Sudarshan Sreeram, Young D. Kwon, Cecilia Mascolo

Test-time adaptation (TTA) offers a compelling remedy for machine learning (ML) models that degrade under domain shifts, improving generalisation on-the-fly with only unlabelled samples. This flexibility suits real deployments, yet conventional evaluations unrealistically assume unbounded processing time, overlooking the accuracy-latency trade-off. As ML increasingly underpins latency-sensitive and user-facing use-cases, temporal pressure constrains the viability of adaptable inference; predictions arriving too late to act on are futile. We introduce Tempora, a framework for evaluating TTA under this pressure. It consists of temporal scenarios that model deployment constraints, evaluation protocols that operationalise measurement, and time-contingent utility metrics that quantify the accuracy-latency trade-off. We instantiate the framework with three such metrics: (1) discrete utility for asynchronous streams with hard deadlines, (2) continuous utility for interactive settings where value decays with latency, and (3) amortised utility for budget-constrained deployments. Applying Tempora to seven TTA methods on ImageNet-C across 240 temporal evaluations reveals rank instability: conventional rankings do not predict rankings under temporal pressure; ETA, a state-of-the-art method in the conventional setting, falls short in 41.2% of evaluations. The highest-utility method varies with corruption type and temporal pressure, with no clear winner. By enabling systematic evaluation across diverse temporal constraints for the first time, Tempora reveals when and why rankings invert, offering practitioners a lens for method selection and researchers a target for deployable adaptation.

Yuwei Zhang, Tong Xia, Jing Han et al.

Respiratory audio, such as coughing and breathing sounds, has predictive power for a wide range of healthcare applications, yet is currently under-explored. The main problem for those applications arises from the difficulty in collecting large labeled task-specific data for model development. Generalizable respiratory acoustic foundation models pretrained with unlabeled data would offer appealing advantages and possibly unlock this impasse. However, given the safety-critical nature of healthcare applications, it is pivotal to also ensure openness and replicability for any proposed foundation model solution. To this end, we introduce OPERA, an OPEn Respiratory Acoustic foundation model pretraining and benchmarking system, as the first approach answering this need. We curate large-scale respiratory audio datasets (~136K samples, over 400 hours), pretrain three pioneering foundation models, and build a benchmark consisting of 19 downstream respiratory health tasks for evaluation. Our pretrained models demonstrate superior performance (against existing acoustic models pretrained with general audio on 16 out of 19 tasks) and generalizability (to unseen datasets and new respiratory audio modalities). This highlights the great promise of respiratory acoustic foundation models and encourages more studies using OPERA as an open resource to accelerate research on respiratory audio for health. The system is accessible from https://github.com/evelyn0414/OPERA.

Tong Xia, Abhirup Ghosh, Xinchi Qiu et al.

Federated Learning (FL) enables model development by leveraging data distributed across numerous edge devices without transferring local data to a central server. However, existing FL methods still face challenges when dealing with scarce and label-skewed data across devices, resulting in local model overfitting and drift, consequently hindering the performance of the global model. In response to these challenges, we propose a pioneering framework called FLea, incorporating the following key components: i) A global feature buffer that stores activation-target pairs shared from multiple clients to support local training. This design mitigates local model drift caused by the absence of certain classes; ii) A feature augmentation approach based on local and global activation mix-ups for local training. This strategy enlarges the training samples, thereby reducing the risk of local overfitting; iii) An obfuscation method to minimize the correlation between intermediate activations and the source data, enhancing the privacy of shared features. To verify the superiority of FLea, we conduct extensive experiments using a wide range of data modalities, simulating different levels of local data scarcity and label skew. The results demonstrate that FLea consistently outperforms state-of-the-art FL counterparts (among 13 of the experimented 18 settings, the improvement is over 5% while concurrently mitigating the privacy vulnerabilities associated with shared features. Code is available at https://github.com/XTxiatong/FLea.git.

Anish Das, Young D. Kwon, Jagmohan Chauhan et al.

Deep Learning (DL) has shown impressive performance in many mobile applications. Most existing works have focused on reducing the computational and resource overheads of running Deep Neural Networks (DNN) inference on resource-constrained mobile devices. However, the other aspect of DNN operations, i.e. training (forward and backward passes) on smartphone GPUs, has received little attention thus far. To this end, we conduct an initial analysis to examine the feasibility of on-device training on smartphones using mobile GPUs. We first employ the open-source mobile DL framework (MNN) and its OpenCL backend for running compute kernels on GPUs. Next, we observed that training on CPUs is much faster than on GPUs and identified two possible bottlenecks related to this observation: (i) computation and (ii) memory bottlenecks. To solve the computation bottleneck, we optimize the OpenCL backend's kernels, showing 2x improvements (40-70 GFLOPs) over CPUs (15-30 GFLOPs) on the Snapdragon 8 series processors. However, we find that the full DNN training is still much slower on GPUs than on CPUs, indicating that memory bottleneck plays a significant role in the lower performance of GPU over CPU. The data movement takes almost 91% of training time due to the low bandwidth. Lastly, based on the findings and failures during our investigation, we present limitations and practical guidelines for future directions.

Yuwei Zhang, Kumar Ayush, Siyuan Qiao et al.

We present SensorLM, a family of sensor-language foundation models that enable wearable sensor data understanding with natural language. Despite its pervasive nature, aligning and interpreting sensor data with language remains challenging due to the lack of paired, richly annotated sensor-text descriptions in uncurated, real-world wearable data. We introduce a hierarchical caption generation pipeline designed to capture statistical, structural, and semantic information from sensor data. This approach enabled the curation of the largest sensor-language dataset to date, comprising over 59.7 million hours of data from more than 103,000 people. Furthermore, SensorLM extends prominent multimodal pretraining architectures (e.g., CLIP, CoCa) and recovers them as specific variants within a generic architecture. Extensive experiments on real-world tasks in human activity analysis and healthcare verify the superior performance of SensorLM over state-of-the-art in zero-shot recognition, few-shot learning, and cross-modal retrieval. SensorLM also demonstrates intriguing capabilities including scaling behaviors, label efficiency, sensor captioning, and zero-shot generalization to unseen tasks.

Yu Wu, Ting Dang, Dimitris Spathis et al. · cambridge

Contrastive learning (CL) has emerged as a promising approach for representation learning in time series data by embedding similar pairs closely while distancing dissimilar ones. However, existing CL methods often introduce false negative pairs (FNPs) by neglecting inherent characteristics and then randomly selecting distinct segments as dissimilar pairs, leading to erroneous representation learning, reduced model performance, and overall inefficiency. To address these issues, we systematically define and categorize FNPs in time series into semantic false negative pairs and temporal false negative pairs for the first time: the former arising from overlooking similarities in label categories, which correlates with similarities in non-stationarity and the latter from neglecting temporal proximity. Moreover, we introduce StatioCL, a novel CL framework that captures non-stationarity and temporal dependency to mitigate both FNPs and rectify the inaccuracies in learned representations. By interpreting and differentiating non-stationary states, which reflect the correlation between trends or temporal dynamics with underlying data patterns, StatioCL effectively captures the semantic characteristics and eliminates semantic FNPs. Simultaneously, StatioCL establishes fine-grained similarity levels based on temporal dependencies to capture varying temporal proximity between segments and to mitigate temporal FNPs. Evaluated on real-world benchmark time series classification datasets, StatioCL demonstrates a substantial improvement over state-of-the-art CL methods, achieving a 2.9% increase in Recall and a 19.2% reduction in FNPs. Most importantly, StatioCL also shows enhanced data efficiency and robustness against label scarcity.

Hong Jia, Young D. Kwon, Dong Ma et al.

Traditional machine learning techniques are prone to generating inaccurate predictions when confronted with shifts in the distribution of data between the training and testing phases. This vulnerability can lead to severe consequences, especially in applications such as mobile healthcare. Uncertainty estimation has the potential to mitigate this issue by assessing the reliability of a model's output. However, existing uncertainty estimation techniques often require substantial computational resources and memory, making them impractical for implementation on microcontrollers (MCUs). This limitation hinders the feasibility of many important on-device wearable event detection (WED) applications, such as heart attack detection. In this paper, we present UR2M, a novel Uncertainty and Resource-aware event detection framework for MCUs. Specifically, we (i) develop an uncertainty-aware WED based on evidential theory for accurate event detection and reliable uncertainty estimation; (ii) introduce a cascade ML framework to achieve efficient model inference via early exits, by sharing shallower model layers among different event models; (iii) optimize the deployment of the model and MCU library for system efficiency. We conducted extensive experiments and compared UR2M to traditional uncertainty baselines using three wearable datasets. Our results demonstrate that UR2M achieves up to 864% faster inference speed, 857% energy-saving for uncertainty estimation, 55% memory saving on two popular MCUs, and a 22% improvement in uncertainty quantification performance. UR2M can be deployed on a wide range of MCUs, significantly expanding real-time and reliable WED applications.

Jialu Tang, Tong Xia, Yuan Lu et al.

Electrocardiogram (ECG) interpretation requires specialized expertise, often involving synthesizing insights from ECG signals with complex clinical queries posed in natural language. The scarcity of labeled ECG data coupled with the diverse nature of clinical inquiries presents a significant challenge for developing robust and adaptable ECG diagnostic systems. This work introduces a novel multimodal meta-learning method for few-shot ECG question answering, addressing the challenge of limited labeled data while leveraging the rich knowledge encoded within large language models (LLMs). Our LLM-agnostic approach integrates a pre-trained ECG encoder with a frozen LLM (e.g., LLaMA and Gemma) via a trainable fusion module, enabling the language model to reason about ECG data and generate clinically meaningful answers. Extensive experiments demonstrate superior generalization to unseen diagnostic tasks compared to supervised baselines, achieving notable performance even with limited ECG leads. For instance, in a 5-way 5-shot setting, our method using LLaMA-3.1-8B achieves an accuracy of 84.6%, 77.3%, and 69.6% on single verify, choose and query question types, respectively. These results highlight the potential of our method to enhance clinical ECG interpretation by combining signal processing with the nuanced language understanding capabilities of LLMs, particularly in data-constrained scenarios.

Chi Ian Tang, Lorena Qendro, Dimitris Spathis et al. · cambridge

Wearable-based Human Activity Recognition (HAR) is a key task in human-centric machine learning due to its fundamental understanding of human behaviours. Due to the dynamic nature of human behaviours, continual learning promises HAR systems that are tailored to users' needs. However, because of the difficulty in collecting labelled data with wearable sensors, existing approaches that focus on supervised continual learning have limited applicability, while unsupervised continual learning methods only handle representation learning while delaying classifier training to a later stage. This work explores the adoption and adaptation of CaSSLe, a continual self-supervised learning model, and Kaizen, a semi-supervised continual learning model that balances representation learning and down-stream classification, for the task of wearable-based HAR. These schemes re-purpose contrastive learning for knowledge retention and, Kaizen combines that with self-training in a unified scheme that can leverage unlabelled and labelled data for continual learning. In addition to comparing state-of-the-art self-supervised continual learning schemes, we further investigated the importance of different loss terms and explored the trade-off between knowledge retention and learning from new tasks. In particular, our extensive evaluation demonstrated that the use of a weighting factor that reflects the ratio between learned and new classes achieves the best overall trade-off in continual learning.

Gaia A. Bertolino, Yuwei Zhang, Tong Xia et al.

Conversational generative AI is rapidly entering healthcare, where general-purpose models must integrate heterogeneous patient signals and support diverse interaction styles while producing clinically meaningful outputs. In respiratory care, non-invasive audio, such as recordings captured via mobile microphones, enables scalable screening and longitudinal monitoring, but the heterogeneity challenge is particularly acute: recordings vary widely across devices, environments, and acquisition protocols, and questions span multiple intents and question formats. Existing biomedical audio-language QA systems are typically monolithic, without any specialization mechanisms for tackling diverse respiratory corpora and query intents. They are also only validated in limited settings, leaving it unclear how reliably they handle the shifts encountered in real-world settings. To address these limitations, we introduce RAMoEA-QA, a hierarchically routed generative model for respiratory audio question answering that unifies multiple question types and supports both discrete and continuous targets within a single multimodal system. RAMoEA-QA applies two-stage conditional specialization: an Audio Mixture-of-Experts routes each recording to a suitable pre-trained audio encoder, and a Language Mixture-of-Adapters selects a LoRA adapter on a shared frozen LLM to match the query intent and answer format. By specializing both acoustic representations and generation behaviour per example, RAMoEA-QA consistently outperforms strong baselines and routing ablations with minimal parameter overhead, improving in-domain test accuracy to 0.72 (vs. 0.61 and 0.67 for state-of-the-art baselines) and exhibiting the strongest generalization for diagnosis under domain, modality, and task shifts.

Hyungjun Yoon, Seungjoo Lee, Yu Yvonne Wu et al.

Electrophysiological (ExG) signals offer valuable insights into human physiology, yet building foundation models that generalize across everyday tasks remains challenging due to two key limitations: (i) insufficient data diversity, as most ExG recordings are collected in controlled labs with bulky, expensive devices; and (ii) task-specific model designs that require tailored processing (i.e., targeted frequency filters) and architectures, which limit generalization across tasks. To address these challenges, we introduce an approach for scalable, task-agnostic ExG monitoring in the wild. We collected 50 hours of unobtrusive free-living ExG data with an earphone-based hardware prototype to narrow the data diversity gap. At the core of our approach is Physiology-informed Multi-band Tokenization (PiMT), which decomposes ExG signals into 12 physiology-informed tokens, followed by a reconstruction task to learn robust representations. This enables adaptive feature recognition across the full frequency spectrum while capturing task-relevant information. Experiments on our new DailySense dataset-the first to enable ExG-based analysis across five human senses-together with four public ExG benchmarks, demonstrate that PiMT consistently outperforms state-of-the-art methods across diverse tasks.

Cynthia Dong, Hong Jia, Young D. Kwon et al.

While there are many advantages to deploying machine learning models on edge devices, the resource constraints of mobile platforms, the dynamic nature of the environment, and differences between the distribution of training versus in-the-wild data make such deployments challenging. Current test-time adaptation methods are often memory-intensive and not designed to be quantization-compatible or deployed on low-resource devices. To address these challenges, we present LeanTTA, a novel backpropagation-free and stateless framework for quantized test-time adaptation tailored to edge devices. Our approach minimizes computational costs by dynamically updating normalization statistics without backpropagation, which frees LeanTTA from the common pitfall of relying on large batches and historical data, making our method robust to realistic deployment scenarios. Our approach is the first to enable further computational gains by combining partial adaptation with quantized module fusion. We validate our framework across sensor modalities, demonstrating significant improvements over state-of-the-art TTA methods, including a 15.7% error reduction, peak memory usage of only 11.2MB for ResNet18, and fast adaptation within an order-of-magnitude of normal inference speeds on-device. LeanTTA provides a robust solution for achieving the right trade offs between accuracy and system efficiency in edge deployments, addressing the unique challenges posed by limited data and varied operational conditions.

Harry Coppock, Alican Akman, Christian Bergler et al.

The COVID-19 pandemic has caused massive humanitarian and economic damage. Teams of scientists from a broad range of disciplines have searched for methods to help governments and communities combat the disease. One avenue from the machine learning field which has been explored is the prospect of a digital mass test which can detect COVID-19 from infected individuals' respiratory sounds. We present a summary of the results from the INTERSPEECH 2021 Computational Paralinguistics Challenges: COVID-19 Cough, (CCS) and COVID-19 Speech, (CSS).

Zahra Tarkhani, Lorena Qendro, Malachy O'Connor Brown et al.

Brain computing interfaces (BCI) are used in a plethora of safety/privacy-critical applications, ranging from healthcare to smart communication and control. Wearable BCI setups typically involve a head-mounted sensor connected to a mobile device, combined with ML-based data processing. Consequently, they are susceptible to a multiplicity of attacks across the hardware, software, and networking stacks used that can leak users' brainwave data or at worst relinquish control of BCI-assisted devices to remote attackers. In this paper, we: (i) analyse the whole-system security and privacy threats to existing wearable BCI products from an operating system and adversarial machine learning perspective; and (ii) introduce Argus, the first information flow control system for wearable BCI applications that mitigates these attacks. Argus' domain-specific design leads to a lightweight implementation on Linux ARM platforms suitable for existing BCI use-cases. Our proof of concept attacks on real-world BCI devices (Muse, NeuroSky, and OpenBCI) led us to discover more than 300 vulnerabilities across the stacks of six major attack vectors. Our evaluation shows Argus is highly effective in tracking sensitive dataflows and restricting these attacks with an acceptable memory and performance overhead (<15%).

Ting Dang, Jing Han, Tong Xia et al.

Recent work has shown the potential of using audio data (eg, cough, breathing, and voice) in the screening for COVID-19. However, these approaches only focus on one-off detection and detect the infection given the current audio sample, but do not monitor disease progression in COVID-19. Limited exploration has been put forward to continuously monitor COVID-19 progression, especially recovery, through longitudinal audio data. Tracking disease progression characteristics could lead to more timely treatment. The primary objective of this study is to explore the potential of longitudinal audio samples over time for COVID-19 progression prediction and, especially, recovery trend prediction using sequential deep learning techniques. Crowdsourced respiratory audio data, including breathing, cough, and voice samples, from 212 individuals over 5-385 days were analyzed. We developed a deep learning-enabled tracking tool using gated recurrent units (GRUs) to detect COVID-19 progression by exploring the audio dynamics of the individuals' historical audio biomarkers. The investigation comprised 2 parts: (1) COVID-19 detection in terms of positive and negative (healthy) tests, and (2) longitudinal disease progression prediction over time in terms of probability of positive tests. The strong performance for COVID-19 detection, yielding an AUROC of 0.79, a sensitivity of 0.75, and a specificity of 0.71 supported the effectiveness of the approach compared to methods that do not leverage longitudinal dynamics. We further examined the predicted disease progression trajectory, displaying high consistency with test results with a correlation of 0.75 in the test cohort and 0.86 in a subset of the test cohort who reported recovery. Our findings suggest that monitoring COVID-19 evolution via longitudinal audio data has potential in the tracking of individuals' disease progression and recovery.

Tong Xia, Jing Han, Cecilia Mascolo

A biosignal is a signal that can be continuously measured from human bodies, such as respiratory sounds, heart activity (ECG), brain waves (EEG), etc, based on which, machine learning models have been developed with very promising performance for automatic disease detection and health status monitoring. However, dataset shift, i.e., data distribution of inference varies from the distribution of the training, is not uncommon for real biosignal-based applications. To improve the robustness, probabilistic models with uncertainty quantification are adapted to capture how reliable a prediction is. Yet, assessing the quality of the estimated uncertainty remains a challenge. In this work, we propose a framework to evaluate the capability of the estimated uncertainty in capturing different types of biosignal dataset shifts with various degrees. In particular, we use three classification tasks based on respiratory sounds and electrocardiography signals to benchmark five representative uncertainty quantification methods. Extensive experiments show that, although Ensemble and Bayesian models could provide relatively better uncertainty estimations under dataset shifts, all tested models fail to meet the promise in trustworthy prediction and model calibration. Our work paves the way for a comprehensive evaluation for any newly developed biosignal classifiers.

Kevalee Shah, Dimitris Spathis, Chi Ian Tang et al.

Vast quantities of person-generated health data (wearables) are collected but the process of annotating to feed to machine learning models is impractical. This paper discusses ways in which self-supervised approaches that use contrastive losses, such as SimCLR and BYOL, previously applied to the vision domain, can be applied to high-dimensional health signals for downstream classification tasks of various diseases spanning sleep, heart, and metabolic conditions. To this end, we adapt the data augmentation step and the overall architecture to suit the temporal nature of the data (wearable traces) and evaluate on 5 downstream tasks by comparing other state-of-the-art methods including supervised learning and an adversarial unsupervised representation learning method. We show that SimCLR outperforms the adversarial method and a fully-supervised method in the majority of the downstream evaluation tasks, and that all self-supervised methods outperform the fully-supervised methods. This work provides a comprehensive benchmark for contrastive methods applied to the wearable time-series domain, showing the promise of task-agnostic representations for downstream clinical outcomes.

Young D. Kwon, Jagmohan Chauhan, Abhishek Kumar et al.

Continual learning approaches help deep neural network models adapt and learn incrementally by trying to solve catastrophic forgetting. However, whether these existing approaches, applied traditionally to image-based tasks, work with the same efficacy to the sequential time series data generated by mobile or embedded sensing systems remains an unanswered question. To address this void, we conduct the first comprehensive empirical study that quantifies the performance of three predominant continual learning schemes (i.e., regularization, replay, and replay with examples) on six datasets from three mobile and embedded sensing applications in a range of scenarios having different learning complexities. More specifically, we implement an end-to-end continual learning framework on edge devices. Then we investigate the generalizability, trade-offs between performance, storage, computational costs, and memory footprint of different continual learning methods. Our findings suggest that replay with exemplars-based schemes such as iCaRL has the best performance trade-offs, even in complex scenarios, at the expense of some storage space (few MBs) for training examples (1% to 5%). We also demonstrate for the first time that it is feasible and practical to run continual learning on-device with a limited memory budget. In particular, the latency on two types of mobile and embedded devices suggests that both incremental learning time (few seconds - 4 minutes) and training time (1 - 75 minutes) across datasets are acceptable, as training could happen on the device when the embedded device is charging thereby ensuring complete data privacy. Finally, we present some guidelines for practitioners who want to apply a continual learning paradigm for mobile sensing tasks.

Ashwin Ahuja, Andrea Ferlini, Cecilia Mascolo

Navigation systems are used daily. While different types of navigation systems exist, inertial navigation systems (INS) have favorable properties for some wearables which, for battery and form factors may not be able to use GPS. Earables (aka ear-worn wearables) are living a momentum both as leisure devices, and sensing and computing platforms. The inherent high signal to noise ratio (SNR) of ear-collected inertial data, due to the vibration dumping of the musculoskeletal system; combined with the fact that people typically wear a pair of earables (one per ear) could offer significant accuracy when tracking head movements, leading to potential improvements for inertial navigation. Hence, in this work, we investigate and propose PilotEar, the first end-to-end earable-based inertial navigation system, achieving an average tracking drift of 0.15 m/s for one earable and 0.11 m/s for two earables.

Andrea Ferlini, Dong Ma, Robert Harle et al.

Human gait is a widely used biometric trait for user identification and recognition. Given the wide-spreading, steady diffusion of ear-worn wearables (Earables) as the new frontier of wearable devices, we investigate the feasibility of earable-based gait identification. Specifically, we look at gait-based identification from the sounds induced by walking and propagated through the musculoskeletal system in the body. Our system, EarGate, leverages an in-ear facing microphone which exploits the earable's occlusion effect to reliably detect the user's gait from inside the ear canal, without impairing the general usage of earphones. With data collected from 31 subjects, we show that EarGate achieves up to 97.26% Balanced Accuracy (BAC) with very low False Acceptance Rate (FAR) and False Rejection Rate (FRR) of 3.23% and 2.25%, respectively. Further, our measurement of power consumption and latency investigates how this gait identification model could live both as a stand-alone or cloud-coupled earable system.

Kayla-Jade Butkow, Ting Dang, Andrea Ferlini et al.

With the soaring adoption of in-ear wearables, the research community has started investigating suitable in-ear heart rate (HR) detection systems. HR is a key physiological marker of cardiovascular health and physical fitness. Continuous and reliable HR monitoring with wearable devices has therefore gained increasing attention in recent years. Existing HR detection systems in wearables mainly rely on photoplethysmography (PPG) sensors, however, these are notorious for poor performance in the presence of human motion. In this work, leveraging the occlusion effect that enhances low-frequency bone-conducted sounds in the ear canal, we investigate for the first time \textit{in-ear audio-based motion-resilient} HR monitoring. We first collected HR-induced sounds in the ear canal leveraging an in-ear microphone under stationary and three different activities (i.e., walking, running, and speaking). Then, we devised a novel deep learning based motion artefact (MA) mitigation framework to denoise the in-ear audio signals, followed by an HR estimation algorithm to extract HR. With data collected from 20 subjects over four activities, we demonstrate that hEARt, our end-to-end approach, achieves a mean absolute error (MAE) of 3.02 $\pm$ 2.97~BPM, 8.12 $\pm$ 6.74~BPM, 11.23 $\pm$ 9.20~BPM and 9.39 $\pm$ 6.97~BPM for stationary, walking, running and speaking, respectively, opening the door to a new non-invasive and affordable HR monitoring with usable performance for daily activities. Not only does hEARt outperform previous in-ear HR monitoring work, but it outperforms reported in-ear PPG performance.

Erika Bondareva, Elín Rós Hauksdóttir, Cecilia Mascolo

Bruxism is a disorder characterised by teeth grinding and clenching, and many bruxism sufferers are not aware of this disorder until their dental health professional notices permanent teeth wear. Stress and anxiety are often listed among contributing factors impacting bruxism exacerbation, which may explain why the COVID-19 pandemic gave rise to a bruxism epidemic. It is essential to develop tools allowing for the early diagnosis of bruxism in an unobtrusive manner. This work explores the feasibility of detecting bruxism-related events using earables in a mimicked in-the-wild setting. Using inertial measurement unit for data collection, we utilise traditional machine learning for teeth grinding and clenching detection. We observe superior performance of models based on gyroscope data, achieving an 88% and 66% accuracy on grinding and clenching activities, respectively, in a controlled environment, and 76% and 73% on grinding and clenching, respectively, in an in-the-wild environment.

Erika Bondareva, Jing Han, William Bradlow et al.

Cardiovascular (CV) diseases are the leading cause of death in the world, and auscultation is typically an essential part of a cardiovascular examination. The ability to diagnose a patient based on their heart sounds is a rather difficult skill to master. Thus, many approaches for automated heart auscultation have been explored. However, most of the previously proposed methods involve a segmentation step, the performance of which drops significantly for high pulse rates or noisy signals. In this work, we propose a novel segmentation-free heart sound classification method. Specifically, we apply discrete wavelet transform to denoise the signal, followed by feature extraction and feature reduction. Then, Support Vector Machines and Deep Neural Networks are utilised for classification. On the PASCAL heart sound dataset our approach showed superior performance compared to others, achieving 81% and 96% precision on normal and murmur classes, respectively. In addition, for the first time, the data were further explored under a user-independent setting, where the proposed method achieved 92% and 86% precision on normal and murmur, demonstrating the potential of enabling automatic murmur detection for practical use.

Lorena Qendro, Alexander Campbell, Pietro Liò et al.

Electroencephalography (EEG) is crucial for the monitoring and diagnosis of brain disorders. However, EEG signals suffer from perturbations caused by non-cerebral artifacts limiting their efficacy. Current artifact detection pipelines are resource-hungry and rely heavily on hand-crafted features. Moreover, these pipelines are deterministic in nature, making them unable to capture predictive uncertainty. We propose E4G, a deep learning framework for high frequency EEG artifact detection. Our framework exploits the early exit paradigm, building an implicit ensemble of models capable of capturing uncertainty. We evaluate our approach on the Temple University Hospital EEG Artifact Corpus (v2.0) achieving state-of-the-art classification results. In addition, E4G provides well-calibrated uncertainty metrics comparable to sampling techniques like Monte Carlo dropout in just a single forward pass. E4G opens the door to uncertainty-aware artifact detection supporting clinicians-in-the-loop frameworks.

Jing Han, Tong Xia, Dimitris Spathis et al.

Researchers have been battling with the question of how we can identify Coronavirus disease (COVID-19) cases efficiently, affordably and at scale. Recent work has shown how audio based approaches, which collect respiratory audio data (cough, breathing and voice) can be used for testing, however there is a lack of exploration of how biases and methodological decisions impact these tools' performance in practice. In this paper, we explore the realistic performance of audio-based digital testing of COVID-19. To investigate this, we collected a large crowdsourced respiratory audio dataset through a mobile app, alongside recent COVID-19 test result and symptoms intended as a ground truth. Within the collected dataset, we selected 5,240 samples from 2,478 participants and split them into different participant-independent sets for model development and validation. Among these, we controlled for potential confounding factors (such as demographics and language). The unbiased model takes features extracted from breathing, coughs, and voice signals as predictors and yields an AUC-ROC of 0.71 (95\% CI: 0.65$-$0.77). We further explore different unbalanced distributions to show how biases and participant splits affect performance. Finally, we discuss how the realistic model presented could be integrated in clinical practice to realize continuous, ubiquitous, sustainable and affordable testing at population scale.

Benjamin Lucas Searle, Dimitris Spathis, Marios Constantinides et al.

Body-focused repetitive behaviors (BFRBs), like face-touching or skin-picking, are hand-driven behaviors which can damage one's appearance, if not identified early and treated. Technology for automatic detection is still under-explored, with few previous works being limited to wearables with single modalities (e.g., motion). Here, we propose a multi-sensory approach combining motion, orientation, and heart rate sensors to detect BFRBs. We conducted a feasibility study in which participants (N=10) were exposed to BFRBs-inducing tasks, and analyzed 380 mins of signals under an extensive evaluation of sensing modalities, cross-validation methods, and observation windows. Our models achieved an AUC > 0.90 in distinguishing BFRBs, which were more evident in observation windows 5 mins prior to the behavior as opposed to 1-min ones. In a follow-up qualitative survey, we found that not only the timing of detection matters but also models need to be context-aware, when designing just-in-time interventions to prevent BFRBs.

Dong Ma, Andrea Ferlini, Cecilia Mascolo

Smart earbuds are recognized as a new wearable platform for personal-scale human motion sensing. However, due to the interference from head movement or background noise, commonly-used modalities (e.g. accelerometer and microphone) fail to reliably detect both intense and light motions. To obviate this, we propose OESense, an acoustic-based in-ear system for general human motion sensing. The core idea behind OESense is the joint use of the occlusion effect (i.e., the enhancement of low-frequency components of bone-conducted sounds in an occluded ear canal) and inward-facing microphone, which naturally boosts the sensing signal and suppresses external interference. We prototype OESense as an earbud and evaluate its performance on three representative applications, i.e., step counting, activity recognition, and hand-to-face gesture interaction. With data collected from 31 subjects, we show that OESense achieves 99.3% step counting recall, 98.3% recognition recall for 5 activities, and 97.0% recall for five tapping gestures on human face, respectively. We also demonstrate that OESense is compatible with earbuds' fundamental functionalities (e.g. music playback and phone calls). In terms of energy, OESense consumes 746 mW during data recording and recognition and it has a response latency of 40.85 ms for gesture recognition. Our analysis indicates such overhead is acceptable and OESense is potential to be integrated into future earbuds.

Young D. Kwon, Jagmohan Chauhan, Cecilia Mascolo

Various incremental learning (IL) approaches have been proposed to help deep learning models learn new tasks/classes continuously without forgetting what was learned previously (i.e., avoid catastrophic forgetting). With the growing number of deployed audio sensing applications that need to dynamically incorporate new tasks and changing input distribution from users, the ability of IL on-device becomes essential for both efficiency and user privacy. However, prior works suffer from high computational costs and storage demands which hinders the deployment of IL on-device. In this work, to overcome these limitations, we develop an end-to-end and on-device IL framework, FastICARL, that incorporates an exemplar-based IL and quantization in the context of audio-based applications. We first employ k-nearest-neighbor to reduce the latency of IL. Then, we jointly utilize a quantization technique to decrease the storage requirements of IL. We implement FastICARL on two types of mobile devices and demonstrate that FastICARL remarkably decreases the IL time up to 78-92% and the storage requirements by 2-4 times without sacrificing its performance. FastICARL enables complete on-device IL, ensuring user privacy as the user data does not need to leave the device.

Sandra Servia-Rodriguez, Cecilia Mascolo, Young D. Kwon

Despite much research targeted at enabling conventional machine learning models to continually learn tasks and data distributions sequentially without forgetting the knowledge acquired, little effort has been devoted to account for more realistic situations where learning some tasks accurately might be more critical than forgetting previous ones. In this paper we propose a Bayesian inference based framework to continually learn a set of real-world, sensing-based analysis tasks that can be tuned to prioritize the remembering of previously learned tasks or the learning of new ones. Our experiments prove the robustness and reliability of the learned models to adapt to the changing sensing environment, and show the suitability of using uncertainty of the predictions to assess their reliability.

Krittika D'Silva, Jordan Cambe, Anastasios Noulas et al.

Modelling the dynamics of urban venues is a challenging task as it is multifaceted in nature. Demand is a function of many complex and nonlinear features such as neighborhood composition, real-time events, and seasonality. Recent advances in Graph Convolutional Networks (GCNs) have had promising results as they build a graphical representation of a system and harness the potential of deep learning architectures. However, there has been limited work using GCNs in a temporal setting to model dynamic dependencies of the network. Further, within the context of urban environments, there has been no prior work using dynamic GCNs to support venue demand analysis and prediction. In this paper, we propose a novel deep learning framework which aims to better model the popularity and growth of urban venues. Using a longitudinal dataset from location technology platform Foursquare, we model individual venues and venue types across London and Paris. First, representing cities as connected networks of venues, we quantify their structure and note a strong community structure in these retail networks, an observation that highlights the interplay of cooperative and competitive forces that emerge in local ecosystems of retail businesses. Next, we present our deep learning architecture which integrates both spatial and topological features into a temporal model which predicts the demand of a venue at the subsequent time-step. Our experiments demonstrate that our model can learn spatio-temporal trends of venue demand and consistently outperform baseline models. Relative to state-of-the-art deep learning models, our model reduces the RSME by ~ 28% in London and ~ 13% in Paris. Our approach highlights the power of complex network measures and GCNs in building prediction models for urban environments. The model could have numerous applications within the retail sector to better model venue demand and growth.

Jordan Cambe, Krittika D'Silva, Anastasios Noulas et al.

Understanding the impact that a new business has on the local market ecosystem is a challenging task as it is multifaceted in nature. Past work in this space has examined the collaborative or competitive role of homogeneous venue types (i.e. the impact of a new bookstore on existing bookstores). However, these prior works have been limited in their scope and explanatory power. To better measure retail performance in a modern city, a model should consider a number of factors that interact synchronously. This paper is the first which considers the multifaceted types of interactions that occur in urban cities when examining the impact of new businesses. We first present a modeling framework which examines the role of new businesses in their respective local areas. Using a longitudinal dataset from location technology platform Foursquare, we model new venue impact across 26 major cities worldwide. Representing cities as connected networks of venues, we quantify their structure and characterise their dynamics over time. We note a strong community structure emerging in these retail networks, an observation that highlights the interplay of cooperative and competitive forces that emerge in local ecosystems of retail establishments. We next devise a data-driven metric that captures the first-order correlation on the impact of a new venue on retailers within its vicinity accounting for both homogeneous and heterogeneous interactions between venue types. Lastly, we build a supervised machine learning model to predict the impact of a given new venue on its local retail ecosystem. Our approach highlights the power of complex network measures in building machine learning prediction models. These models have numerous applications within the retail sector and can support policymakers, business owners, and urban planners in the development of models to characterize and predict changes in urban settings.

Tong Xia, Jing Han, Lorena Qendro et al.

Recently, sound-based COVID-19 detection studies have shown great promise to achieve scalable and prompt digital pre-screening. However, there are still two unsolved issues hindering the practice. First, collected datasets for model training are often imbalanced, with a considerably smaller proportion of users tested positive, making it harder to learn representative and robust features. Second, deep learning models are generally overconfident in their predictions. Clinically, false predictions aggravate healthcare costs. Estimation of the uncertainty of screening would aid this. To handle these issues, we propose an ensemble framework where multiple deep learning models for sound-based COVID-19 detection are developed from different but balanced subsets from original data. As such, data are utilized more effectively compared to traditional up-sampling and down-sampling approaches: an AUC of 0.74 with a sensitivity of 0.68 and a specificity of 0.69 is achieved. Simultaneously, we estimate uncertainty from the disagreement across multiple models. It is shown that false predictions often yield higher uncertainty, enabling us to suggest the users with certainty higher than a threshold to repeat the audio test on their phones or to take clinical tests if digital diagnosis still fails. This study paves the way for a more robust sound-based COVID-19 automated screening system.

Björn W. Schuller, Anton Batliner, Christian Bergler et al.

The INTERSPEECH 2021 Computational Paralinguistics Challenge addresses four different problems for the first time in a research competition under well-defined conditions: In the COVID-19 Cough and COVID-19 Speech Sub-Challenges, a binary classification on COVID-19 infection has to be made based on coughing sounds and speech; in the Escalation SubChallenge, a three-way assessment of the level of escalation in a dialogue is featured; and in the Primates Sub-Challenge, four species vs background need to be classified. We describe the Sub-Challenges, baseline feature extraction, and classifiers based on the 'usual' COMPARE and BoAW features as well as deep unsupervised representation learning using the AuDeep toolkit, and deep feature extraction from pre-trained CNNs using the Deep Spectrum toolkit; in addition, we add deep end-to-end sequential modelling, and partially linguistic analysis.

Chi Ian Tang, Ignacio Perez-Pozuelo, Dimitris Spathis et al.

Machine learning and deep learning have shown great promise in mobile sensing applications, including Human Activity Recognition. However, the performance of such models in real-world settings largely depends on the availability of large datasets that captures diverse behaviors. Recently, studies in computer vision and natural language processing have shown that leveraging massive amounts of unlabeled data enables performance on par with state-of-the-art supervised models. In this work, we present SelfHAR, a semi-supervised model that effectively learns to leverage unlabeled mobile sensing datasets to complement small labeled datasets. Our approach combines teacher-student self-training, which distills the knowledge of unlabeled and labeled datasets while allowing for data augmentation, and multi-task self-supervision, which learns robust signal-level representations by predicting distorted versions of the input. We evaluated SelfHAR on various HAR datasets and showed state-of-the-art performance over supervised and previous semi-supervised approaches, with up to 12% increase in F1 score using the same number of model parameters at inference. Furthermore, SelfHAR is data-efficient, reaching similar performance using up to 10 times less labeled data compared to supervised approaches. Our work not only achieves state-of-the-art performance in a diverse set of HAR datasets, but also sheds light on how pre-training tasks may affect downstream performance.

Lorena Qendro, Jagmohan Chauhan, Alberto Gil C. P. Ramos et al.

Neural networks (NNs) lack measures of "reliability" estimation that would enable reasoning over their predictions. Despite the vital importance, especially in areas of human well-being and health, state-of-the-art uncertainty estimation techniques are computationally expensive when applied to resource-constrained devices. We propose an efficient framework for predictive uncertainty estimation in NNs deployed on embedded edge systems with no need for fine-tuning or re-training strategies. To meet the energy and latency requirements of these embedded platforms the framework is built from the ground up to provide predictive uncertainty based only on one forward pass and a negligible amount of additional matrix multiplications with theoretically proven correctness. Our aim is to enable already trained deep learning models to generate uncertainty estimates on resource-limited devices at inference time focusing on classification tasks. This framework is founded on theoretical developments casting dropout training as approximate inference in Bayesian NNs. Our layerwise distribution approximation to the convolution layer cascades through the network, providing uncertainty estimates in one single run which ensures minimal overhead, especially compared with uncertainty techniques that require multiple forwards passes and an equal linear rise in energy and latency requirements making them unsuitable in practice. We demonstrate that it yields better performance and flexibility over previous work based on multilayer perceptrons to obtain uncertainty estimates. Our evaluation with mobile applications datasets shows that our approach not only obtains robust and accurate uncertainty estimations but also outperforms state-of-the-art methods in terms of systems performance, reducing energy consumption (up to 28x), keeping the memory overhead at a minimum while still improving accuracy (up to 16%).

Jing Han, Chloë Brown, Jagmohan Chauhan et al.

The development of fast and accurate screening tools, which could facilitate testing and prevent more costly clinical tests, is key to the current pandemic of COVID-19. In this context, some initial work shows promise in detecting diagnostic signals of COVID-19 from audio sounds. In this paper, we propose a voice-based framework to automatically detect individuals who have tested positive for COVID-19. We evaluate the performance of the proposed framework on a subset of data crowdsourced from our app, containing 828 samples from 343 participants. By combining voice signals and reported symptoms, an AUC of $0.79$ has been attained, with a sensitivity of $0.68$ and a specificity of $0.82$. We hope that this study opens the door to rapid, low-cost, and convenient pre-screening tools to automatically detect the disease.