Courses of Academic Year 2017/2018


Data Series Management

Lecturer: Themis Palpanas, Paris Descartes University
Download the lecturer CV

Schedule:

  • Mer 2 Maggio: 14-16 P2.5 --- 16-18 P0.4
  • Gio 3 Maggio: 11-13 P1.1 --- 16-18 P1.3
  • Ven 4 Maggio: 09-13 P1.1
  • Lun 7 Maggio: 16-19 P1.5

Program:

  • Data series representations: formal definitions, normalization, summarizations (dft, dct, dwt, paa, pla, apca, isax), amnesic summarizations, experimental comparison of summarizations
  • Data series distance functions: Euclidean, pearson correlation, dtw, constrained dtw, lcss, lower bounds, tightness and pruning power, uniform scaling
  • Data series similarity queries: definitions, nearest neighbor, epsilon-range, serial scan, best so far, early abandoning
  • Data series indexing: R-Trees, indexing for dtw, indexing for uniform scaling, isax, isax2+, ads+, vertical scan, experimental comparison of indexes
  • Data series mining: classification, clustering, motifs, discords, matrix profile
  • Streaming data series processing: summarizations for streaming data series (paa, pla, dwt, dft), similarity matching (wedgie, spring, ssm), fast serial scan (Euclidean, dtw)
  • Using Matlab for data series processing

Exam:
Small project

CFD: 6


Deep Learning for Fault Prediction

Lecturer: Prof. Roberto Paredes Palacios
Associate Professor at Departamento de Sistemas Informáticos y Computación DSIC of the Universidad Poliécnica de Valencia UPV. I belong to the Pattern Recognition and Human Language Technologies Research Center PRHLT Roberto Paredes is the Director of the PRHLT and the CTO and founder of Solver Machine Learning

Schedule:

MO27 Piano Terra - Sala Master

  • Mon 12 Feb, 14:00-18:00 (4 hours)
  • Tue 13 Feb, 10:00-14:00 (2 hours + 2 hours Lab)
  • Wed 14 Feb, 10:00-14:00 (2 hours + 2 hours Lab)
  • Thu 15 Feb, 09:00-12:00 (3 hours)

Program:

Deep Learning can be seen as an evolution of Neural Networks to deal with the Representation Learning problem. The use of deep learning provides a way to tackle the recognition problems end-to-end. The same neural network model provides a representation of the input signals to recognize (images, speech, ...) and a seamless classification model.

In these lectures we will learn from the basic fully connected models to the convolutional networks. Important details regarding regularization, normalization or data augmentation will be considered as well. Some basic code will be provide in order to practically apply the theoretical concepts.

Finally some applications will be mentioned:

  • Fault prediction in factories
  • Fraud detection in Visa transactions
  • Logo detection
  • Energy price forecasting
  • Face recognition


Theoretical Requirements: Basic Machine Learning, Linear Algebra, Basic Image Processing
Practical Requirements: Python Programming

Exam:
Test with closed questions

CFD: 5


Big Data Management & Analytics in collaboration with CINECA

Lecturer: to be defined

Schedule:
to be defined

Program:
to be defined

Exam:
Not yet defined

CFD: 6


Academic English Course

Lecturer: to be defined

Schedule:
to be defined

Program:
Basic English grammar for academic research.

Exam:
Not yet defined

CFD: 3