На 2-ри юли (вторник) от 14:00 ч. в зала 503 на ИМИ ще се състои сбирка на
Семинара по Изследване на операциите.
Доклад на тема
On some contemporary mathematical methods for data analysis
ще изнесе проф. Пандо Георгиев, ИМИ-БАН.
Абстракт:
We will review some selected methods and algorithms for data analysis:
A) Latent component analysis – reveals hidden features of composite data. Particularly, we will describe some
Signal Separation Methods:
- based on statistical independence – Independent Component Analysis
- based on sparsity. – Sparse Component Analysis
- based on non-negativity – Non-Negative Matrix and Tensor Factorization
B) New approach to multiclass learning problems and applications to some problems in machine learning, as:
- nonlinear least square problems
- subspace clustering and non-linear skeletons
- reconstruction of several functions from mixed samples
- support vector kernel regression (under development)
C) kernel methods for data analysis
D) tensor decompositions and tensor networks.
Some of the above topics are interconnected and subject to intensive development with numerous applications.