Academic Year 2016/2017 - 1° Year - Curriculum Data Science
Teaching Staff: Giuseppe Nicosia
Credit Value: 6
Scientific field: INF/01 - Informatics
Taught classes: 24 hours
Term / Semester:

Learning Objectives

Knowledge and understanding: Students will acquire basic knowledge about them most important models of Machine Learning and algorithm techniques for automatic learning from data.


Applying knowledge and understanding: students will be to able to apply the acquired knowledge in several fields such as: recognition and classification, social networking, computer visione and natural language understanding.
Autonomia di giudizio (making judgements): Students will be able to evaluate the possibility of developing algorithms and machine learning systems in different application fields.
Communication skills: students will acquire the necessary communication skills and appropriate linguistic skills to explain and clarify problems relative to machine learning systems and their applications.
Capacità di apprendimento (learning skills): students will be able to adapt the acquire knowledge to new contexts as well and to understand the limits of applicability of machine learning techniques

Detailed Course Content

Detailed course content:

1. Machine Learning Basics

1.1 Learning Algorithms

1.2 Capacity, Overfitting and Underfitting

1.3 Hyperparameters and Validation Sets

1.4 Estimators, Bias and Variance

1.5 Maximum Likelihood Estimation

1.6 Bayesian Statistics

1.7 Supervised Learning Algorithms

1.8 Unsupervised Learning Algorithms

1.9 Stochastic Gradient Descent

1.10 Building a Machine Learning Algorithm

1.11 Challenges Motivating Deep Learning

2. Deep Feedforward Networks

3. Regularization for Deep Learning

4. Optimization for Training Deep Models

5. Convolutional Networks

6. Sequence Modeling: Recurrent and Recursive Nets

7. Practical Methodology and Applications

8. Linear Factor Models

9. Autoencoders

10. Representation Learning

11. Structured Probabilistic Models for Deep Learning

12. Approximate Inference

13. Deep Generative Models

13.1 Boltzmann Machines

13.2 Restricted Boltzmann Machines

13.3 Deep Belief Networks

13.4 Deep Boltzmann Machines

13.5 Boltzmann Machines for Real-Valued Data

13.6 Convolutional Boltzmann Machines

13.7 Boltzmann Machines for Structured or Sequential Outputs

13.8 Other Boltzmann Machines

13.9 Back-Propagation through Random Operations

13.10 Directed Generative Nets

13.11 Drawing Samples from Autoencoders

13.12 Generative Stochastic Networks

13.13 Other Generation Schemes

13.14 Evaluating Generative Models

14. Monte Carlo Methods & Confronting the Partition Function

Textbook Information

Required textbook are

Deep Learning, Ian Goodfellow, Yoshua Bengio, Aaron Courville, MIT Press.

Machine Learning: A Probabilistic Perspective, K. P. Murphy, MIT Press, 2012.

Pattern Recognition and Machine Learning, Christopher M. Bishop, Springer Nature.

Other material will be provided by the instructor in class.