You are here: Home » Events » Machine Learning for Signal Processing

Machine Learning for Signal Processing

There are many interesting problems that can be addressed by signal processing. With the advent of ever increasing computational power, there are some problems that were either not possible to address before or not in the way that it can be addressed now. Separate developments have taken place in learning algorithms for over more than the last thirty years. Much more recently interesting and innovative researches have been taking place in the area of Machine Learning for Signal Processing (MLSP). In these lectures, I wish to tell you about some of these efforts, relating to some of my own experiences.

SPEAKER

Prof. Asoke Nandi, David Jardine Chair of Signal Processing
The University of Liverpool
Liverpool, U.K

PRELIMINARY SCHEDULE
Thursday 29th of October (room "Brattain")
09:00 Morning coffee
09:15 Opening
09.30 Introduction to machine learning
11:00 Lunch break
12:00 Feature Selection methods
13:30 Coffee break
13:45 Classification methods
15:15 Break
15:30 Case studies on breast cancer detection

Friday 30th of October (in the morning room "Brattain", after lunch changing to "Big Seminar Hall", "Iso seminaarisali")
8:30 Morning coffee
8:45 Case studies on rotating machine condition monitoring
10:15 Break
10:30 Case studies on automatic modulation recognition
12:00 Lunch
13:00 Feature generation using genetic programming
14:30 Coffee
14:45 Comparative partner selection and code-bloat in genetic programming, independent component analysis, and other applications.

EVENT DESCRIPTION

Lecture 1 will be a general introduction to machine learning, including detection, classification and recognition; it will also outline the various stages of the process.

Lectures 2 and 3 will cover feature selection methods and classification methods.

Lecture 4 will describe case studies with breast cancer data.

Lecture 5 will report case studies on rotating machine condition monitoring.

Lecture 6 will outline case studies in automatic modulation recognition, a central element in software defined radio.

Lecture 7 will present more recent ideas of feature generation using genetic programming and record their performance with breast cancer data, rotating machine vibration data, and audio data.

Lecture 8 will discuss comparative partner selection and code-bloat in genetic programming, independent component analysis, and other applications.

REGISTRATION

Please, registrate through our website . Participation fee for one participation is 360e. For INFORTE member organisations' staff, participation is free-of-charge. Meals and accommodation are not included.

For PhD students it is possible to gain 2-3 credit points when participating this course actively.