Machine Learning

• Weekly hours
• Objectives
• Contents
• Activities
• Teaching methodology
• Evaluation methodology
• Bibliography
• Previous capacities

Teachers

Person in charge

• Alexandre Perera Lluna ( )

Weekly hours

Objectives

1. Apply deep learning models to biological and biomedical data, selecting and adapting architectures such as convolutional, recurrent, and transformer networks to solve specific problems in bioinformatics and bioengineering.
   Preprocess, represent, and analyze heterogeneous biomedical data (medical images, biological sequences, physiological signals), using modern computational tools and good scientific programming practices.
   Interpret and evaluate the performance of deep learning models in biomedical contexts, using appropriate metrics and understanding the limitations and ethical risks of using artificial intelligence in biomedicine.
   Related competences: K2, K3, S2, S3, S4, C6,
2. Preprocess, represent, and analyze heterogeneous biomedical data (medical images, biological sequences, physiological signals), using modern computational tools and good scientific programming practices.
   Related competences: K4, K5, S4, S8, C3, C6,
3. Interpret and evaluate the performance of deep learning models in biomedical contexts, using appropriate metrics and understanding the limitations and ethical risks of using artificial intelligence in biomedicine.
   Related competences: K4, K5, S4, S8, C3,

Contents

1. Introduction to Deep Learning in Bioinformatics & Bioengineering
   Introduction to Deep Learning in Bioinformatics & Bioengineering, Biological Data Types and software tools for effective learning workflows.
2. Multilayer Perceptrons: Foundations of Deep Learning
   An introduction to the fundamental building blocks of deep learning through multilayer perceptrons (MLPs), covering network architecture, activation functions, forward and backward propagation, and their role in modeling complex nonlinear relationships in biological data.
3. Biomedical Image Analysis with CNNs
   This module explores convolutional neural networks (CNNs) for analyzing biomedical images, focusing on feature extraction, classification, and segmentation tasks in applications such as pathology, radiology, and microscopy.
4. Biological Time-Series and Sequence Modeling with RNNs
   An in-depth look at recurrent neural networks (RNNs) and their variants for modeling sequential and temporal biological data, including time series, physiological signals, and nucleotide or protein sequences.
5. Transformers and Attention Mechanisms for Biological Sequences
   Introduction to the attention mechanisms and transformer architectures, emphasizing their effectiveness in capturing long-range dependencies in biological sequences for tasks such as structure prediction, and functional annotation
6. Encoder-Decoder Architectures
   Design and application of encoder-decoder models for transforming, annotating, and generating biological sequences and biomedical data, with emphasis on RNN and transformer variants, attention integration, and real-world sequence modeling tasks.
7. Language Models and NLP in Bioinformatics and Biomedicine
   This module explores how language models are applied to biomedical texts and biological sequences, enabling tasks such as named entity recognition, document classification, and functional annotation through pretrained models like BioBERT, DNABERT, and protein sequence transformers.

Activities

Activity Evaluation act

Teaching methodology

Evaluation methodology

Bibliography

Basic:

• Deep Learning for the Life Sciences - Bharath Ramsundar, Peter Eastman, Pat Walters, Vijay Pande, O'Reilly Media, Inc., ISBN: 9781492039839
  
• Bioinformatics with Python Cookbook - Tiago Antão, ISBN: 978-1789344691
  

Previous capacities