Self-tracking apps are a prominent component of today’s wellness and health management strategies. These apps collect a variety of personal data, which is typically stored in a remote central location and utilised by machine learning models. Federated Learning (FL) brings about a fundamental change in this for machine learning. Through interviews and one workshop with 18 university students (majority of whom (16) identified as women), we investigate whether users’ increased awareness of enhanced-privacy protection in exchange for potentially reduced accuracy and fairness afforded by FL can alter users’ perceptions of these apps and the types of data they feel comfortable sharing. Participants’ willingness to pay for the FL application varied depending on their sensitivity towards data privacy. However, participants were still reluctant to share certain types of data even after knowing that FL keeps data on their devices. This was due to a persistent lack of trust in companies accompanied by a lack of awareness of the privacy implications of sharing sensitive data. Our findings suggest the need to further inform and educate users about the privacy advantages of FL. Companies must also prioritise establishing trust with users, as this was found to be a strong factor in increasing users’ acceptability of the apps, irrespective of the privacy-enhanced technology employed.

Reinforcement learning (RL) holds promise for supporting personalised decision-making in healthcare, but existing approaches often struggle to incorporate patient expertise and individual preferences, key components for clinically viable AI systems. This work introduces PAINT (Preference Adaptation for Individualised Treatment), a general framework for preference-guided offline RL in safety-critical settings. PAINT combines sketch-based reward annotation with safety-constrained policy optimisation, enabling fine-grained preference capture from historical patient data without requiring action labels. A reward model trained on this feedback guides offline RL while supporting tunable sensitivity to preference signals and enforcing clinical safety constraints. Using type 1 diabetes (T1D) management as a case study, in-silico evaluation with the FDA-accepted T1D simulator demonstrates that can PAINT reduces patient risk by 15% over commercial baselines under guidance, while enabling preference-driven adaptations such as improved management during challenging mealtime events and enhanced robustness to dosing errors. The method further shows resilience to real-world challenges including sample size, annotation noise, and inter-patient variability. These findings suggest PAINT offers a practical pathway for integrating human feedback into offline RL in patient settings, with broader implications for developing trustworthy and adaptive AI systems in healthcare.

With the rise of social media, fake news has been known to spread rapidly through the platforms and significantly influencing public opinion. However, traditional machine learning techniques are difficult in addressing fake news detection problem due to the highly imbalanced distribution of classes. As an alternative, reinforcement learning (RL) can be used to address the imbalanced prediction problem by designing different reward functions. However, it ignores the dynamic and adversarial environment in social networks, where malicious user can influence the platform’s decision. To address these challenges, we propose a multi-agent reinforcement learning (MARL) framework called Information Classification Markov Game (ICMG) to model fake news detection as a competitive game between a malicious user and a content moderator. Within this framework, we explore the best reward function for a content moderator to achieve better performance in a real world dataset. Additionally, we compare the effectiveness of different types of agents including those based on randomness, Minimax and Q-Learning. Our results show that a MARL setting with both players using Q-Learning setting achieves the best performance. Compared with the single-agent framework, ICMG effectively improves macro F1 score, improving the model performance on the real-world dataset. Moreover, this framework effectively simulates the adversarial dynamics in real-world social media platforms, providing a new approach for developing more realistic content moderation systems.

With the development and growing use of social media platforms in the last two decades, platform architectures have driven how we notice, consume and share information. Whilst centralised social networks and the use of recommendation algorithms are a prominently used architecture, in recent years an alternative and novel framework has emerged aiming to offer users a non-commercial decentralised platform to distribute content. Run by users of the platform, Mastodon offers many of the benefits of traditional centralised approaches, however, with the absence of recommendation algorithms there is risk that these architectures could instead promote echo-chambers and the growth of disinformation. With this in mind, we collect a new large Mastodon dataset, consisting of three million connections between over a hundred thousand users. Modelling content using 68 conversational features, and measuring influence using twelve different metrics, we analyse the most common topics being discussed between influential users, the conversational features present in influential content, and the relationships between influence measurements. Our analysis finds a strong correlation between influence and negative traits at every network resolution, with positive and neutral traits in some cases being negatively correlated with influence. Our analysis also shows that influential users have a strong relationship with social/political commentary.

Parkinson's disease (PD) is a slowly progressive, debilitating neurodegenerative disease which causes motor symptoms including gait dysfunction. Motor fluctuations are alterations between periods with a positive response to levodopa therapy ("on") and periods marked by re-emergency of PD symptoms ("off") as the response to medication wears off. These fluctuations often affect gait speed and they increase in their disabling impact as PD progresses. To improve the effectiveness of current indoor localisation methods, a transformer-based approach utilising dual modalities which provide complementary views of movement, Received Signal Strength Indicator (RSSI) and accelerometer data from wearable devices, is proposed. A sub-objective aims to evaluate whether indoor localisation, including its in-home gait speed features (i.e. the time taken to walk between rooms), could be used to evaluate motor fluctuations by detecting whether the person with PD is taking levodopa medications or withholding them. To properly evaluate our proposed method, we use a free-living dataset where the movements and mobility are greatly varied and unstructured as expected in real-world conditions. 24 participants lived in pairs (consisting of one person with PD, one control) for five days in a smart home with various sensors. Our evaluation on the resulting dataset demonstrates that our proposed network outperforms other methods for indoor localisation. The sub-objective evaluation shows that precise room-level localisation predictions, transformed into in-home gait speed features, produce accurate predictions on whether the PD participant is taking or withholding their medications.

The widespread adoption of effective hybrid closed loop systems would represent an important milestone of care for people living with type 1 diabetes (T1D). These devices typically utilise simple control algorithms to select the optimal insulin dose for maintaining blood glucose levels within a healthy range. Online reinforcement learning (RL) has been utilised as a method for further enhancing glucose control in these devices. Previous approaches have been shown to reduce patient risk and improve time spent in the target range when compared to classical control algorithms, but are prone to instability in the learning process, often resulting in the selection of unsafe actions. This work presents an evaluation of offline RL for developing effective dosing policies without the need for potentially dangerous patient interaction during training. This paper examines the utility of BCQ, CQL and TD3-BC in managing the blood glucose of the 30 virtual patients available within the FDA-approved UVA/Padova glucose dynamics simulator. When trained on less than a tenth of the total training samples required by online RL to achieve stable performance, this work shows that offline RL can significantly increase time in the healthy blood glucose range from 61.6±0.3% to 65.3±0.5% when compared to the strongest state-of-art baseline (p<0.001). This is achieved without any associated increase in low blood glucose events. Offline RL is also shown to be able to correct for common and challenging control scenarios such as incorrect bolus dosing, irregular meal timings and compression errors. The code for this work is available at: https://github.com/hemerson1/offline-glucose.

In interdisciplinary spaces such as digital health, datasets that are complex to collect, require specialist facilities, and/or are collected with specific populations have value in a range of different sectors. In this study we collected a simulated free-living dataset, in a smart home, with 12 participants (six people with Parkinson’s, six carers). We explored their initial perceptions of the sensors through interviews and then conducted two data exploration workshops, wherein we showed participants the collected data and discussed their views on how this data, and other data relating to their Parkinson’s symptoms, might be shared across different sectors. We provide recommendations around how participants might be better engaged in considering data sharing in the early stages of research, and guidance for how research might be configured to allow for more informed data sharing practices in the future.

The feasibility of integrating the temporal nature of the Bluetooth Low Energy (BLE) Received Signal Strength Indicator (RSSI) into a self-supervised machine learning model for room-level and sub-room-level localization in a realistic residential setting is investigated. The signal is transmitted by a wearable wrist watch and received by multiple access points acting as receivers communicating via the BLE standard. It is found that while the baseline room-level accuracy is sufficiently high for practical applications in rooms separated by a wall, confusion can occur between non-adjacent rooms and thus lower localization performance. Two approaches are explored that exploit the time dimension of the data to mitigate this problem: maximum likelihood estimation in a conditional random field model, and self-supervised contrastive learning based on temporal proximity. Using a real world dataset collected in residential homes, we develop the approaches on the data collected in one residence before evaluating them on data collected in another. On the evaluation residence, we find that conditional random fields do not improve upon the baseline in terms of the weighted F1 score, while contrastive learning leads to an improvement in localization performance.

Online social network platforms have a problem with misinformation. One popular way of addressing this problem is via the use of machine learning based automated misinformation detection systems to classify if a post is misinformation. Instead of post hoc detection, we propose to predict if a user will engage with misinformation in advance and design an effective graph neural network classifier based on ego-graphs for this task. However, social networks are highly dynamic, reflecting continual changes in user behaviour, as well as the content being posted. This is problematic for machine learning models which are typically trained on a static training dataset, and can thus become outdated when the social network changes. Inspired by the success of continual learning on such problems, we propose an ego-graphs replay strategy in continual learning (EgoCL) using graph neural networks to effectively address this issue. We have evaluated the performance of our method on user engagement with misinformation on two Twitter datasets across nineteen misinformation and conspiracy topics. Our experimental results show that our approach EgoCL has better performance in terms of predictive accuracy and computational resources than the state of the art.

Misinformation is becoming increasingly prevalent on social media and in news articles. It has become so widespread that we require algorithmic assistance utilising machine learning to detect such content. Training these machine learning models require datasets of sufficient scale, diversity and quality. However, datasets in the field of automatic misinformation detection are predominantly monolingual, include a limited amount of modalities and are not of sufficient scale and quality. Addressing this, we develop a data collection and linking system (MuMiN-trawl), to build a public misinformation graph dataset (MuMiN), containing rich social media data (tweets, replies, users, images, articles, hashtags) spanning 21 million tweets belonging to 26 thousand Twitter threads, each of which have been semantically linked to 13 thousand fact-checked claims across dozens of topics, events and domains, in 41 different languages, spanning more than a decade. The dataset is made available as a heterogeneous graph via a Python package (mumin). We provide baseline results for two node classification tasks related to the veracity of a claim involving social media, and demonstrate that these are challenging tasks, with the highest macro-average F1-score being 62.55% and 61.45% for the two tasks, respectively. The MuMiN ecosystem is available at https://mumin-dataset.github.io/, including the data, documentation, tutorials and leaderboards.

Data augmentation has been widely used in machine learning for natural language processing and computer vision tasks to improve model performance. However, little research has studied data augmentation on graph neural networks, particularly using augmentation at both train- and test-time. Inspired by the success of augmentation in other domains, we have designed a method for social influence prediction using graph neural networks with train- and test-time augmentation, which can effectively generate multiple augmented graphs for social networks by utilising a variational graph autoencoder in both scenarios. We have evaluated the performance of our method on predicting user influence on multiple social network datasets. Our experimental results show that our end-to-end approach, which jointly trains a graph autoencoder and social influence behaviour classification network, can outperform state-of-the-art approaches, demonstrating the effectiveness of train-and test-time augmentation on graph neural networks for social influence prediction. We observe that this is particularly effective on smaller graphs.

Edge computing is rapidly changing the IoT-Cloud landscape. Various testbeds are now able to run multiple Docker-like containers developed and deployed by end-users on edge devices. However, this capability may allow an attacker to deploy a malicious container on the host and compromise it. This paper presents a dataset based on the Linux Auditing System, which contains malicious and benign container activity. We developed two malicious scenarios, a denial of service and a privilege escalation attack, where an adversary uses a container to compromise the edge device. Furthermore, we deployed benign user containers to run in parallel with the malicious containers. Container activity can be captured through the host system via system calls. Our time series auditd dataset contains partial labels for the benign and malicious related system calls. Generating the dataset is largely automated using a provided AutoCES framework. We also present a semi-supervised machine learning use case with the collected data to demonstrate its utility. The dataset and framework code are open-source and publicly available.

This paper presents Vesta, a digital health platform composed of a smart home in a box for data collection and a machine learning based analytic system for deriving health indicators using activity recognition, sleep analysis and indoor localization. This system has been deployed in the homes of 40 patients undergoing a heart valve intervention in the United Kingdom (UK) as part of the EurValve project, measuring patients health and well-being before and after their operation. In this work a cohort of 20 patients are analyzed, and 2 patients are analyzed in detail as example case studies. A quantitative evaluation of the platform is provided using patient collected data, as well as a comparison using standardized Patient Reported Outcome Measures (PROMs) which are commonly used in hospitals, and a custom survey. It is shown how the ubiquitous in-home Vesta platform can increase clinical confidence in self-reported patient feedback. Demonstrating its suitability for digital health studies, Vesta provides deeper insight into the health, well-being and recovery of patients within their home.

Wireless localisation is becoming increasingly important in various applications such as smart homes, elderly healthcare facilities and in industry where centimetre (cm) level localisation accuracy is desired. Ultra-wideband (UWB) systems can be used for such applications since they can achieve a ranging precision below 10 cm in a Line-of-Sight (LoS) setup. However, in non LoS(NLoS) scenarios these systems provide a lower accuracy. In this paper, we exploit the high resolution Channel Impulse Response (CIR) provided by the Decawave EVK1000 boards for localisation in a residential environment. We employ a single anchor node and use the CIR obtained from five different locations as fingerprints to investigate whether the location of the tag can be accurately estimated in NLoS scenarios. Our investigation showed that the CIR can be effectively used as fingerprints to provide a location classification accuracy as high as 98% when the environment remains relatively stable. However, using the CIR data recorded in a second experiment (same setup as first experiment) as test data and applying the trained model of the first experiment to it showed a significant degradation in performance (50-60% accuracy) due to the changes in the environment. On the other hand, by using five features extracted from the UWB signals for location classification, an accuracy in excess of 99% is obtained during testing in both experiments.

Traditionally, the vision community has devised algorithms to estimate the distance between an original image and images that have been subject to perturbations. Inspiration was usually taken from the human visual perceptual system and how the system processes different perturbations in order to replicate to what extent it determines our ability to judge image quality. While recent works have presented deep neural networks trained to predict human perceptual quality, very few borrow any intuitions from the human visual system. To address this, we present PerceptNet, a convolutional neural network where the architecture has been chosen to reflect the structure and various stages in the human visual system. We evaluate PerceptNet on various traditional perception datasets and note strong performance on a number of them as compared with traditional image quality metrics. We also show that including a nonlinearity inspired by the human visual system in classical deep neural networks architectures can increase their ability to judge perceptual similarity. Compared to similar deep learning methods, the performance is similar, although our network has a number of parameters that is several orders of magnitude less.

Passive wireless sensing using WiFi signals has become a very active area of research over the past few years. Such techniques provide a cost-effective and non-intrusive solution for human activity sensing especially in healthcare applications. One of the main approaches used in wireless sensing is based on fine-grained WiFi Channel State Information (CSI) which can be extracted from commercial Network Interface Cards (NICs). In this paper, we present a new signal processing pipeline required for effective wireless sensing. An experiment involving five participants performing six different activities was carried out in an office space to evaluate the performance of activity recognition using WiFi CSI in different physical layouts. Experimental results show that the CSI system has the best detection performance when activities are performed half-way in between the transmitter and receiver in a line-of-sight (LoS) setting. In this case, an accuracy as high as 91% is achieved while the accuracy for the case where the transmitter and receiver are co-located is around 62%. As for the case when data from all layouts is combined, which better reflects the real-world scenario, the accuracy is around 67%. The results showed that the activity detection performance is dependent not only on the locations of the transmitter and receiver but also on the positioning of the person performing the activity.

Deep clustering has increasingly been demonstrating superiority over conventional shallow clustering algorithms. Deep clustering algorithms usually combine representation learning with deep neural networks to achieve this performance, typically optimizing a clustering and non-clustering loss. In such cases, an autoencoder is typically connected with a clustering network, and the final clustering is jointly learned by both the autoencoder and clustering network. Instead, we propose to learn an autoencoded embedding and then search this further for the underlying manifold. For simplicity, we then cluster this with a shallow clustering algorithm, rather than a deeper network. We study a number of local and global manifold learning methods on both the raw data and autoencoded embedding, concluding that UMAP in our framework is able to find the best clusterable manifold of the embedding. This suggests that local manifold learning on an autoencoded embedding is effective for discovering higher quality clusters. We quantitatively show across a range of image and time-series datasets that our method has competitive performance against the latest deep clustering algorithms, including outperforming current state-of-the-art on several. We postulate that these results show a promising research direction for deep clustering. The code can be found at https://github.com/rymc/n2d.

Reciprocal Recommender Systems are recommender systems for social platforms that connect people to people. They are commonly used in online dating, social networks and recruitment services. The main difference between these and conventional user-item recommenders that might be found on, for example, a shopping service, is that they must consider the interests of both parties. In this study, we present a novel method of making reciprocal recommendations based on image data. Given a user’s history of positive and negative preference expressions on other users images, we train a siamese network to identify images that fit a user’s personal preferences. We provide an algorithm to interpret those individual preference indicators into a single reciprocal preference relation. Our evaluation was performed on a large real-world dataset provided by a popular online dating service. Based on this, our service significantly improves on previous state-of-the-art content-based solutions, and also out-performs collaborative filtering solutions in cold-start situations. The success of this model provides empirical evidence for the high importance of images in online dating.

Most traditional data mining algorithms struggle to cope with the sheer scale of data efficiently. In this paper, we propose a general framework to accelerate existing algorithms to cluster large-scale datasets which contain large numbers of attributes, items, and clusters. Our framework makes use of locality sensitive hashing to significantly reduce the cluster search space. We also theoretically prove that our framework has a guaranteed error bound in terms of the clustering quality. This framework can be applied to a set of centroid-based clustering algorithms that assign an object to the most similar cluster, and we adopt the popular K-Modes categorical clustering algorithm to present how the framework can be applied. We validated our framework with five synthetic datasets and a real world Yahoo! Answers dataset. The experimental results demonstrate that our framework is able to speed up the existing clustering algorithm between factors of 2 and 6, while maintaining comparable cluster purity.

In this paper we propose a graph stream clustering algorithm with a unified similarity measure on both structural and attribute properties of vertices, with each attribute being treated as a vertex. Unlike others, our approach does not require an input parameter for the number of clusters, instead, it dynamically creates new sketch-based clusters and periodically merges existing similar clusters. Experiments on two publicly available datasets reveal the advantages of our approach in detecting vertex clusters in the graph stream. We provide a detailed investigation into how parameters affect the algorithm performance. We also provide a quantitative evaluation and comparison with a well-known offline community detection algorithm which shows that our streaming algorithm can achieve comparable or better average cluster purity.