The Thesis was developed in partnership with company C-Side.

Abstract:

Detecting anomalies in time series data is an important task in areas such as energy, healthcare and security. The progress made in anomaly detection has been mostly based on approaches using supervised machine learning algorithms that require big labelled datasets to be trained. However, in the context of applications, collecting and annotating such large-scale datasets is difficult, time-consuming or even too expensive, while it requires domain knowledge from experts in the field. Therefore, anomaly detection has been such a great challenge for researchers and practitioners. This Thesis proposes a generic, unsupervised and scalable framework for anomaly detection in time series data. The proposed approach is based on a variational autoencoder, a deep generative model that combines variational inference with deep learning. Moreover, the architecture integrates recurrent neural networks to capture the sequential nature of time series data and its temporal dependencies. Furthermore, an attention mechanism is introduced to improve the performance of the encoding-decoding process. The results on solar energy generation and electrocardiogram time series data show the ability of the proposed model to detect anomalous patterns in time series from different fields of application, while providing structured and expressive data representations.

[Thesis, Slides]

Designed and implemented a deep learning approach for fake video detection. This problem was solved as a supervised video classification task with an architecture based on recurrent neural networks and convolutional neural networks.

This project was developed in the framework of a challenge promoted by the United Nations UNICRI Centre for Artificial Intelligence and Robotics, in the Hackathon for Peace, Justice, and Security.

This project was developed under the framework of energy transition. It was proposed an approach to segment and classify satellite images, using convolutional neural networks, and those results were then used to assess the solar potential of various regions in the Netherlands.

The objective of the project was to analyze the material lifecycle of parts within the worldwide ASML supply chain. Process mining techniques were applied to identify the underlying processes and extract insights from those.

Designed a recommendation system based on low-rank matrix factorization for company TE Connectivity.

Condition-based recoiling. The objective was to find a strategy to execute predictive maintenance of a complex industrial system based on sensor data.

Proposed an approach for load time series forecasting based on neural networks. The objective was to forecast, at substation level, the load peak, in order to better plan and optimize the smart grid.

Proposed an unsupervised approach for monitoring ECG sequences. (See the Publications section for further details)

Exception identification in water discharge data for damage control.