Skip to content

42055 - Perceptual Learning (AP) [UB]

Type: S3 Course
Semester: Spring
Teaching Points: 15
Offer: Annual
Responsible Unit: UB
Responsible: Laura Igual (UB)
Language: English


In this course, the main aspects of perceptual learning will be reviewed, using visual data as general example. The course will go through its principal steps: feature extraction, classification and validation.
After an introduction to perceptual learning, dimensionality reduction will be presented as a way for characterizing images. The most useful classification methods for computer vision problems will be presented and the role of loss function and regularizer terms in some particular problem solutions will be anlayzed. Moreover, different methods of optimization will be explained, related to the previously reviewed problems. Finally, useful statistical validation methods will be validated. Students will be guided to develop a project focus on visual learning.

See detailed current content here.

1. Introduction to Perceptual Learning.
a. Feature space for describing visual data
b. Perceptual learning pipeline: description of steps
c. General concepts of perceptual data

2. Dimensionality reduction in feature extraction
a. Principal Component Analysis
b. Linear Discriminant Analysis
c. Independent Component Analysis
d. Computer vision examples: eigenfaces, eigenmotions

3. Classification
a. Kernel-based learning in computer vision: linear and non-linear kernel
b. Application of ensemble learning: Bagging, Boosting
c. Review of application regularizer terms in computer vision problems
d. Review of optimization methods
e. Definition of different loss functions for computer vision problems.

4. Statistical Validation
a. Hypothesis testing, statistical significance
b. Cross-validation and Leave-one-out validation methods
c. Parametric and non-parametric tests