Machine Learning: Neural Networks and Advanced Models
Code 321AA
Credits 6
Learning outcomes
Machine Learning: Neural Networks and Advanced Models
Objectives
The course provide the methodologies needed to specialize in the area of design of new advanced machine learning models, including state-of-the-art neural networks, considering the processing of complex domains and non-vectorial data. The paradigm of dynamical neural networks is the ground to introduce the adaptive processing of sequences and variable-size structures. The course focus also on the critical analysis of the synergy between methodological developments and the design of innovative interdisciplinary applications on Natural Science complex domains, and on the introduction to research topics.
Syllabus
- Neural Networks, advanced aspects:
Historical notes, Regularization, Constructive approaches, Neurocomputing approaches for unsupervised learning.
- Models for Structured Data:
- Structured domains and learning tasks for sequences, time series and graphs.
- Dynamical Recurrent Neural Networks: architectures, learning algorithm, properties.
- Generative approaches: Hidden Markov Models.
- Recursive models for supervised and unsupervised learning.
- Kernel-based approaches for complex (non-vectorial) data.
- Emerging approaches for structured domains and relational learning.
- Applications for applicative and interdisciplinary science: Case-studies
in Bioinformatics and Cheminformatics.
- Emerging topics in the machine learning research area.
Objectives
The course provide the methodologies needed to specialize in the area of design of new advanced machine learning models, including state-of-the-art neural networks, considering the processing of complex domains and non-vectorial data. The paradigm of dynamical neural networks is the ground to introduce the adaptive processing of sequences and variable-size structures. The course focus also on the critical analysis of the synergy between methodological developments and the design of innovative interdisciplinary applications on Natural Science complex domains, and on the introduction to research topics.
Syllabus
- Neural Networks, advanced aspects:
Historical notes, Regularization, Constructive approaches, Neurocomputing approaches for unsupervised learning.
- Models for Structured Data:
- Structured domains and learning tasks for sequences, time series and graphs.
- Dynamical Recurrent Neural Networks: architectures, learning algorithm, properties.
- Generative approaches: Hidden Markov Models.
- Recursive models for supervised and unsupervised learning.
- Kernel-based approaches for complex (non-vectorial) data.
- Emerging approaches for structured domains and relational learning.
- Applications for applicative and interdisciplinary science: Case-studies
in Bioinformatics and Cheminformatics.
- Emerging topics in the machine learning research area.