The course is aimed at providing the theoretical and applicative basis of Pattern Recognition techniques for the computational analysis of biological objects with a complex structure (such as graphs, sequences, networks, strings and so on). In particular, the course introduces and discusses the most important computational techniques for the analysis of structured data, with particular emphasis on the representation and on the generative and discriminative approaches.
At the end of the course, the students will be able to analyse a biological problem, involving complex and structured biological data, from a Pattern Recognition perspective; the will also have the skills needed to study, invent, develop and implement the different components of a Pattern Recognition System for biological structured data.
CHAPTER 1 Basic Pattern Recognition concepts and introduction to structured data
CHAPTER 2. Representation of structured data
- Advanced dimensionality reduction techniques
- The Bag of words representation
- The dissimilarity-based representation
CHAPTER 3. Models for structured data
- Generative models
- Bayes Networks
- Learning and inference
CHAPTER 4. Kernels for structured data
- Support Vector Machines e kernel
- Kernels for structured data
CHAPTER 5. Deep Learning
R. Duda, P. Hart, D. Stork Pattern Classification. Wiley, 2001
P. Baldi, S. Brunak, Bioinformatics, The Machine Learning Approach. MIT Press, 2001
C.M. Bishop, Pattern Recognition and Machine Learning, Springer, 2006
The exam is aimed at the verification of the following skills:
- capability of clearly and concisely describe the different components of a Pattern Recognition System for structured data
- capability of analise, understand and describe a Pattern Recognition system (or a given part of it) relative to a biological problem which involves structured data
The exam consists of two parts
i) a written exam containing questions on topics presented during the course (15 points available). The written part is passed is the grade is greater or equal to 8.
ii) an oral presentation of a scientific paper published in relevant bioinformatics journals or conferences on a given argument (decided during the course). The paper is chosen by the candidate and approved by the instructor (15 points available).
The two parts of the exam can be passed separately: the final grade is the sum of the two grades.
The total exam is passed if the final grade is greater or equal to 18. Each evaluation is maintained valid for the whole academic year.