| Upcoming Events -Check It Out | NIPS
2003 Workshop
on Sparse Representations |
| |
 |
Justinian Rosca
Ph.D. - Computer Science,
University of Rochester
Program Manager - Audio & Signal Processing
Solutions
Siemens
Corporate Research
Room D221
Multimedia and Video
Technology
755 College Road East ,
Princeton, New Jersey 08540, USA
Voice (609) 734-3365
Fax (609) 734-6565
Email: justinian.rosca siemens.com
|
Epistle:
"When, at night, with drooping eyelids, I blow out the candles flare
Time's unending path is followed only by the old clock there;
For just draw aside the curtains and the moon will flood the room
With a fire of passions summoned by the arduous of her gloom;
From the night of recollection she will resurrect an eon
Of distress - which we, however, sense as in a dream-like paean."
Mihai Eminescu |
|
Since joining Siemens Corporate Research, I have been interested in the blind source separation
problem and generally in audio and statistical signal processing applications.
I consider it fascinating to disentangle sounds and in particular speech from a
virtually incomprehensible mixture of voice signals and noise. Practical
solutions to such difficult problems raise interesting theoretical issues
related to the representation of signals (henceforth interest in sparse
representations) and models appropriate for statistical estimation.
Computational learning and machine learning as well stochastic search techniques
are a continuation of interest during graduate school days. My Ph.D.
dissertation dealt with the the scaling problem in learning, methods for
coupling learning and simulated evolution, the study of emergence of
self-organization, global optimization and search, and adaptive principles.
Application areas included learning structure (e.g. modular neural networks and
Bayesian networks), signal analysis and feature discovery, knowledge discovery.
My enthusiasm was modeled "naturally" by my dissertation
advisor, Prof. Dana
Ballard, and the effect persists :-)
|
| |
The research presented at this Web Site has been
sponsored by Siemens Corporate
Research (1998-2003) and National Science Foundation Grants
(1992-1197). |
| © Justinian Rosca 2003 |
[ Home ] [ Talks/Events ] [ Publications ] [ Photos ] [ Demos ]
|
|