Machine learning has become a fundamental component of numerous commercial and research applications. Using the Python programming language and libraries such as PyTorch, it is possible to rapidly develop sophisticated applications in areas including audio-based human–computer interaction (e.g., natural language dialogue systems), music information retrieval, and many others.

With this in mind, the course aims to introduce students to the field of machine learning. Within this framework, students will study the principles underlying the various stages involved in implementing data-driven knowledge discovery systems, using both classical machine learning techniques and state-of-the-art methods. The course also covers the latest approaches to natural language interaction based on large language models. Lectures will address the complete machine learning workflow, including data collection, feature extraction, model training, and performance evaluation.

In addition to providing the necessary theoretical background, the course makes extensive use of Python libraries that are widely employed in both research and commercial applications for the development of automatic pattern recognition systems in domains such as natural language processing, audio signal processing, and related fields.

Upon successful completion of the course, students will be able to:

Understand the fundamental concepts and applications of machine learning.

Evaluate the advantages and limitations of widely used machine learning algorithms.

Design datasets and construct reliable training and evaluation sets for data-driven knowledge discovery.

Apply advanced model evaluation techniques and hyperparameter optimization methods.

Demonstrate proficiency in the use of popular Python-based machine learning libraries and frameworks.

Employ trained large language models to develop natural language human–computer interaction systems.

The following topics will be covered in the Applied Machine Learning course.

Theory

Introduction to machine learning: linear regression and logistic classification

Handwritten digit classification from images: introduction to neural networks

Convolutional neural networks for image and audio processing; transposed (deconvolutional) neural networks

Autoencoders, Variational Autoencoders (VAEs), and Generative Adversarial Networks (GANs)

Semantic embedding methods for time-series data

Time-series modeling and processing using Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks

Transformers and Large Language Models (LLMs)

Large Language Models for tool use, Retrieval-Augmented Generation (RAG), and Agentic AI

Laboratory Exercises

Forward propagation in neural networks

Handwritten digit classification

Convolutional neural networks for image and audio processing

Variational Autoencoders for audio applications

Semantic word embedding exercises for natural language

Using LSTM networks for music emotion recognition

Using Large Language Models for natural language interaction

Using Large Language Models and external tools for music analysis