Tutorials
The role of the tutorials is to provide a platform for a more intensive scientific exchange amongst researchers interested in a particular topic and as a meeting point for the community. Tutorials complement the depth-oriented technical sessions by providing participants with broad overviews of emerging fields. A tutorial can be scheduled for 1.5 or 3 hours.
Tutorial proposals are accepted until:
January 23, 2026
If you wish to propose a new Tutorial please kindly fill out and submit this
Expression of Interest form.
Tutorial on
A Hands-on Introduction to Time Series Feature Extraction
Instructor
|
Duarte Folgado
Fraunhofer Portugal AICOS
Portugal
|
|
|
|
Brief Bio
Duarte Folgado is a Senior Scientist at the Intelligent Systems research group at Fraunhofer AICOS. He studied Biomedical Engineering and completed his MSc (2015) and PhD (2023) at the NOVA School of Science and Technology (FCT NOVA). He was a Visiting Graduate Student in 2023 at the Massachusetts Institute of Technology, collaborating with the Institute of Medical Engineering & Science and the MIT.nano. He is also an Invited Assistant Lecturer in the Physics Department of FCT NOVA since 2020. He received the Best Student Award in Biomedical Engineering (2015), the Merit Scholarship Grant (2015), and a Fulbright Award for Research (2022). Currently, he is developing artificial intelligence solutions for healthcare and manufacturing. His main research interests include data mining, machine learning, deep learning, and Explainable AI, specializing in techniques for time series datasets and also in human-AI interaction.
|
Abstract
Abstract
— Are you extracting all the relevant information from your time series data?
Time series are a fundamental data type for understanding the behavior of real-world systems across several domains in data science. This hands-on tutorial supported with code examples will provide an accessible overview of the recent research in time series classification, with a strong emphasis on the task of feature extraction applied to physiological time series. We will use the Time Series Feature Extraction Library (TSFEL) that computes over 65 different features across the statistical, temporal, spectral, and fractal domains. Alongside a brief theoretical introduction to the feature sets, we will cover important practical recommendations for their successful use with biosignal data.
Keywords
Time Series, Feature Extraction, Machine Learning, Python
Aims and Learning Objectives
To understand the basic and intermediate aspects of time series feature extraction applied to physiological time series data.
Target Audience
A hands-on tutorial session of 1.5 hours, including a short lecture-style introduction.
Prerequisite Knowledge of Audience
Basic level of familiarity with Python.
Detailed Outline
1. A general overview of time series classification
2. Feature extraction in physiological time series - introduction to statistical, temporal, spectral, and fractal feature sets
3. Hands-on introduction with the Time Series Feature Extraction Library (TSFEL)
4. Wrap-up and closing remarks.
Bibliography: https://www.sciencedirect.com/science/article/pii/S2352711020300017
Tutorial on
NeuralLib: a Modular, Efficient, Generalizable Approach for Biosignal Processing
Instructor
|
Nianfei Ao
FCT
Portugal
|
|
|
|
Brief Bio
Nianfei Ao is a doctoral candidate in the MSCA PATTERN project, working on the development of AI tools to accurately assess and mitigate the influence of electromagnetic interference on biosignals, with a current focus on denoising methods. Before starting his PhD, he gained hands-on engineering experience in both academia and industry, including building deep learning models for automated analysis of pathology images during his Master’s research and developing software for automatic processing of protein electrophoresis gel images in the big data intelligent analysis department at GenScript. His current work focuses on designing robust, efficient machine learning pipelines for complex biological data and translating them into practical tools for biosignal monitoring and related applications.
|
Abstract
— Are you fully leveraging deep learning for your biosignal data in a sustainable and reusable way?
Deep learning is increasingly used for biosignal analysis, but many workflows still rely on ad-hoc models that are difficult to reuse, compare and deploy efficiently. This hands-on tutorial introduces NeuLib, an open-source deep learning framework for biosignal processing built on three principles: modularity, efficiency and generalization. Through concise code examples, we will show how to select biosignal-specific GRU- and Transformer-based architectures, train and evaluate models, and adapt pre-trained models to new datasets using transfer learning while monitoring computational footprint.
[Note] NeuLib is jointly developed and maintained by Mariana Dias, Nianfei Ao and Hugo Gamboa.
?
Keywords
Biosignals, ECG, Modularity, Deep Learning, Python
Aims and Learning Objectives
This tutorial aims to provide participants with a practical and conceptual introduction to NeuLib, an open-source deep learning framework for biosignal processing built around three principles: modularity, efficiency and generalization. At the end of the tutorial, attendees will be able to:
1). Understand the design goals and core abstractions of NeuLib;
2). Train new models from scratch using standardized training, validation and testing pipelines;
3). Load and reuse pre‑trained models from the NeuLib model hub, and apply model‑based transfer learning to new biosignal datasets with minimal additional training;
4). Select and configure biosignal‑specific neural architectures
Target Audience
A hands-on tutorial session of 1 - 1.5 hours, including a short lecture-style introduction.
Prerequisite Knowledge of Audience
Basic knowledge of machine learning and deep learning (e.g., understanding of neural networks, training/validation/testing).
Detailed Outline
1.Introduction and Motivation on NeuLib
2.NeuLib Architecture and Core Components
3.Hands‑on Session I: Training and Evaluating a Model
4.Hands‑on Session II: Model Reuse and Transfer Learning