Significant advancements have been made in applying machine learning techniques to the classification of surface electromyography (sEMG) signals. Despite these improvements, there remains considerable room for enhancing these methods to achieve even greater accuracy. This is particularly evident in complex sEMG classification problems, where the patterns across different classes can be very similar. This similarity often leads to reduced accuracy in both validation and test sets, highlighting the need for more refined approaches. This study aims to improve sEMG-based hand gesture recognition by introducing a novel Stacked One-vs-One (SOvO) classification approach. Traditional One-vs-One (OvO) strategies decompose multi-class problems into binary classifications but often suffer from cumulative errors of weak classifiers. The proposed SOvO method enhances classification by stacking the probability outputs from pairwise binary classifiers in the OvO approach on the training data as additional features for the final classification. Depending on whether the final classifier is a deep learning or conventional model, different feature integration strategies were explored. A comparative analysis was conducted on the Ninapro DB4, DB5, and DB6 datasets across two window sizes. Results show that SOvO consistently outperforms OvO in accuracy by 3% to 10%. This improvement is statistically significant, as confirmed by a repeated measures ANOVA ( p < 0.001). These findings demonstrate that SOvO provides a more robust and accurate framework for sEMG-based gesture recognition, addressing limitations of conventional OvO methods.

Abstract This study evaluated a smart technology intervention in the home as a support for individuals with severe mental illness. This study recruited 13 participants in a variety of community-based homes. Participants were offered a smartphone, a touchscreen monitor and health devices such as smartwatches, weigh-scales, and automated medication dispensers. Data was exported to the Lawson Integrated DataBase for care providers to monitor/track. Interviews with participants and focus groups with participants and care providers were conducted at baseline, 6-months and 12-months, and survey instruments were used to collect quantitative data about different dimensions of health and social determinants. Descriptive statistics from these outcome measures are presented as the sample size was too small for meaningful statistical inference. Qualitative analyses revealed a high degree of acceptability of the devices and motivation for healthy living, communication and mental health. Health Care Providers also noted improvements to client health. This study proves the feasibility of deploying smart technologies to support individuals with severe mental illness. Future scale-up would further our understanding of their impacts.

Smartphones present unique data collection opportunities for research. Ecological momentary assessment (EMA) studies have been conducted using smartphones in many areas, as participant responses can be collected at the same time as device sensor data. We wanted to study undergraduate mental health using smartphones while investigating data from as many smartphone sensors as possible. The app used for this had to be accessible from the standard app stores to use existing participant phones. The software also had to have an easy setup process at the server-side due to limited resources. As a result, we developed the Ecological Momentary Assessment eXtensions platform. We describe the platform, two studies that used Version 2 of it, and summarize data collection results. We then outline a pathway to a unified EMAX3 app. To the best of our knowledge, EMAX is the only platform developed and used for sensor-based EMA research at a Canadian university.

The overall objective of this research is to evaluate the use of a mobile health smartphone application (app) to improve the mental health of youth between the ages of 14–25 years, with symptoms of anxiety/depression. This project includes 115 youth who are accessing outpatient mental health services at one of three hospitals and two community agencies. The youth and care providers are using eHealth technology to enhance care. The technology uses mobile questionnaires to help promote self-assessment and track changes to support the plan of care. The technology also allows secure virtual treatment visits that youth can participate in through mobile devices. This longitudinal study uses participatory action research with mixed methods. The majority of participants identified themselves as Caucasian (66.9%). Expectedly, the demographics revealed that Anxiety Disorders and Mood Disorders were highly prevalent within the sample (71.9% and 67.5% respectively). Findings from the qualitative summary established that both staff and youth found the software and platform beneficial.

We present new methodology based on Multi-Objective Markov Decision Processes for developing sequential decision support systems from data. Our approach uses sequential decision-making data to provide support that is useful to many different decision-makers, each with different, potentially time-varying preference. To accomplish this, we develop an extension of fitted-$Q$ iteration for multiple objectives that computes policies for all scalarization functions, i.e. preference functions, simultaneously from continuous-state, finite-horizon data. We identify and address several conceptual and computational challenges along the way, and we introduce a new solution concept that is appropriate when different actions have similar expected outcomes. Finally, we demonstrate an application of our method using data from the Clinical Antipsychotic Trials of Intervention Effectiveness and show that our approach offers decision-makers increased choice by a larger class of optimal policies.

In this paper, we present a framework which enables medical decision making in the presence of partial information. At its core is ontology-based automated reasoning, machine learning techniques are integrated to enhance existing patient datasets in order to address the issue of missing data. Our approach supports interoperability between different health information systems. This is clarified in a sample implementation that combines three separate datasets (patient data, drug-drug interactions and drug prescription rules) to demonstrate the effectiveness of our algorithms in producing effective medical decisions. In short, we demonstrate the potential for machine learning to support a task where there is a critical need from medical professionals by coping with missing or noisy patient data and enabling the use of multiple medical datasets.

Fitted value iteration (FVI) with ordinary least squares regression is known to diverge. We present a new method, "Expansion-Constrained Ordinary Least Squares" (ECOLS), that produces a linear approximation but also guarantees convergence when used with FVI. To ensure convergence, we constrain the least squares regression operator to be a non-expansion in the infinity-norm. We show that the space of function approximators that satisfy this constraint is more rich than the space of "averagers," we prove a minimax property of the ECOLS residual error, and we give an efficient algorithm for computing the coefficients of ECOLS based on constraint generation. We illustrate the algorithmic convergence of FVI with ECOLS in a suite of experiments, and discuss its properties.

Recently, we have introduced a novel approach to dynamic programming and re- inforcement learning that is based on maintaining explicit representations of sta- tionary distributions instead of value functions. In this paper, we investigate the convergence properties of these dual algorithms both theoretically and empirically, and show how they can be scaled up by incorporating function approximation.

We present an efficient "sparse sampling" technique for approximating Bayes optimal decision making in reinforcement learning, addressing the well known exploration versus exploitation tradeoff. Our approach combines sparse sampling with Bayesian exploration to achieve improved decision making while controlling computational cost. The idea is to grow a sparse lookahead tree, intelligently, by exploiting information in a Bayesian posterior---rather than enumerate action branches (standard sparse sampling) or compensate myopically (value of perfect information). The outcome is a flexible, practical technique for improving action selection in simple reinforcement learning scenarios.