I am João Pereira. I am currently a PDEng Trainee in Data Science at Eindhoven University of Technology (TU/e). Most of the time I do machine learning and I am broadly interested in the field of data science.
Prior to joining TU/e as a PDEng Trainee, I took a M.Sc. degree in Electrical and Computer Engineering at Instituto Superior Técnico (IST). Meanwhile, I developed research on deep learning applications, within the Signal and Image Processing Group at the Institute for Systems and Robotics (ISR-Lisbon).
My research was mainly focused on deep learning for sequential data and to investigate how deep generative models (such as variational autoencoders), recurrent neural networks and attention mechanisms can be successfuly used for time series data modeling and, in particular, used to solve anomaly detection tasks. In this framework, I worked with energy and healthcare data.
If you want to know more about my work I invite you to check out the Projects section. Please feel free to discuss and share any comments about them with me ;)
The PDEng Program Data Science is a joint initiative of Eindhoven University of Technology (TU/e), Tilburg University (TiU) and the Data Science Centre Eindhoven (DSC/e). Together they founded the Jheronimus Academy of Data Science (JADS). The academy is a thriving data science community, in which postgraduate education, innovation, entrepreneurial activities, lifelong learning and scientific research are integrated. A major part of the PDEng program Data Science is related to real-life, cross-functional projects with partners in the data science ecosystem.
Developed research on deep learning applications.
Developed a data science project on load forecasting using neural networks.
Designed and projected medium/low voltage networks. Smart grid devices installation and maintenance.
Courses: Artificial Intelligence, Robotics, Optimization and Algorithms, Digital Signal Processing, Computer Control, State Space Control, Modeling and Simulation, Signals and Systems, Introduction to Research in Electrical and Computer Engineering.
Courses: Machine Learning, Image Processing and Vision, Instrumentation and Sensors, Electric Power Systems.
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.
Unsupervised anomaly detection in solar PV generation time series data. Propose an approach based on a variational autoencoder (VAE) with encoder and decoder parametrized by recurrent neural networks (Bi-LSTMs). Introduce Variational Self-Attention Mechanism (VSAM).
Conference Paper: Unsupervised Anomaly Detection in Energy Time Series Data using Variational Recurrent Autoencoders with Attention
Oral presentation @ 17th International Conference on Machine Learning and Applications (ICMLA'18)
Orlando, Florida, USA
Keywords: Representations Learning, Variational Autoencoder, Recurrent Neural Networks, Wasserstein Distance, Clustering.
Conference Paper: Learning Representations from Healthcare Time Series Data for Unsupervised Anomaly Detection
Oral presentation @ 6th IEEE International Conference on Big Data and Smart Computing (BigComp'19)