Pervasive Sensing

In this paper, we aim to introduce a novel method for explaining time-series classification, leveraging the capabilities of ChatGPT to enhance the interpretability of results and foster a deeper understanding of feature contributions within time-series data.

Real-time monitoring of human activities using wearable devices often requires the deployment of machine learning models on resource-constrained edge devices. Stateof- the-art Human Activity Recognition neural network models, while effective, already suffer from excessive size and complexity. Furthermore, our systematic analysis reveals that, even worse, the computational cost and model size of most SOTA HAR models escalate significantly with an increase in sensor channels. In this work, we propose an integrated neural architecture search framework to further lighten HAR models. To counteract the negative effects of scaling at deployment time, the proposed framework simultaneously selects and reduces the number of sensor channels, prunes filters, and decreases the temporal dimensions while training the model on optimized hardware.

Tiny edge systems used in IoT devices, wearables or smart textiles are characterized by the need of processing complex sensor data streams under various device constraints. Due to the high number of constraints and the complexity of the optimization of the hyper-parameter space for machine learning based processing, genetic algorithms (GAs) seem to be a perfect fit to enable AutoML for those embedded devices. We introduce a novel Genetic Algorithm (GA) customized for AutoML tasks, addressing the unique challenges posed by highly embedded machine learning domains.

An essential task in predictive maintenance is the prediction of the Remaining Useful Life (RUL) through the analysis of multivariate time series. Using the sliding window method, Convolutional Neural Network (CNN) and conventional Recurrent Neural Network (RNN) approaches have produced impressive results on this matter, due to their ability to learn optimized features. To exploit the multi-level temporal information, many approaches are proposed which combine CNN and RNN models. In this work, we propose a new LSTM variant called embedded convolutional LSTM (ECLSTM).