VITA
ASHUTOSH GARG
Work Address: ****,
Beckman Institute,
Urbana,
IL61801
Home Address: **** *. **** **. *** #108
Urbana,
IL61802
Phone No: 217-***-****
Fax No: 217-***-****
Email: abpnn9@r.postjobfree.com
Web Page:
http://www.ifp.uiuc.edu/~ashutosh
RESEARCH INTERESTS
Multimedia and Human Computer InteractionVideo AnalysisAudio-Visual Speech Detection/RecognitionActivity Detection and ModelingEmotion/Expression RecognitionDriver Work Load managementMachine LearningData dependent Generalization BoundsProbabilistic ClassifiersLearning in Cognitive SituationsOnline learning and learning with unlabeled and
noisy dataBioInformaticsGene Annotation
EDUCATION
2000-2002
PhD in Electrical
and Computer Engineering, University of Illinios at Urbana-Champaign, IL
Dissertation
title: Learning in High dimensional spaces: Applications, Theory and
Algorithms.
Advisor:
Thomas S. Huang; Co-Advisor: Dan Roth
1998-2000 M.S. in Electrical and Computer
Engineering, University of Illinios at Urbana-Champaign, IL
Thesis title:
Multimodal Speaker detection using dynamic Bayesian networks.
Advisor:
Thomas S. Huang
1993-1997
B. Tech in
Electrical Engineering, Indian Institute of Technology, New Delhi, India
Thesis title: Gesture Based
Remote Visualization.
Advisors: Santanu Choudhury and
Subhasish Banerjee
PROFESSIONAL EXPERIENCE
Jun 1998 –
present Research Assistant
Beckman
Institute for Advance Science and Technology
Advisors:
Prof. Thomas S. Huang and Prof. Dan Roth
Working
on Probabilistic and statistical Learning techniques including SVMs, Bayesian
Networks, HMM and Daynamic Bayesian networks with applications to video
analysis, activity recognition and speech recognition.
Developed
theoretical understanding of many learning algorithms in a unified framework
(PAC based, VC-dimension based and probabilistic algorithms)
May 2002-Aug 2002 Research Intern
IBM TJ
Watson Research Lab, Yorktown, NY
Advisors:
Dr. Chalapathy Neti and Dr. Gerasimos Potamianos.
Developed
an audio-visual speech recognizer.
Introduced a variation of Factorial HMM for modeling the frame level
reliability indicators in a multistream HMM framework for AVSR. Developed a
theoretical justification of stream weights using mutual information criterion
and showed that improved performance can be obtained for the task of AVSR by
using frame level reliability indicators.
May 2001-Aug 2001 Research Intern
Microsoft
Research Lab, Seattle, WA
Advisors:
Dr. Eric Horvitz and Dr. Nuria Oliver
Developed
a Hierarchical framework for detecting events in an office scenario. Hidden
Markov Models are used in discriminative fashion to model temporal events
(based on multiple modalities – audio, video and pc activity) to detect
events at different levels of temporal abstraction. The system was presented as
part of the Bill Gates keynote address at the International Joint Conference of
Artificial Intelligence (IJCAI’01). Results published at ICMI’02,
CVPR’01
May 2000-Aug 2000 Research Intern
Cambridge Research Lab (HP Computer Corp.), Cambridge, MA
Advisors: Prof. Vladimir Pavlovic and Prof. Simon Kasif
Annotation
of the Human genome (chromosome 22) and Drosophilae (fruit fly) data using mixture
of experts framework and input/output Hidden Markov Model. A novel I/OHMM based
mixture of experts model is developed that can be used to combine the output of
the temporal sequences to improve the classification performance. Presented at
Computational Genomics’00 and Journal of Bio-Informatics.
May 1999-Aug 1999 Research Intern
Cambridge Research Lab (HP Computer Corp.), Cambridge, MA
Advisors: Prof. Vladimir Pavlovic and Prof. Jim Rehg
Developed
a Multimodal Speaker Detection system using Dynamic Bayesian Networks.
Introduced the framework of “Error feedback DBN” for fusing the
information from different modalities. Excellent results for the task of
speaker detection were obtained using this model. Presented at CVPR’00,
ICMI’00, FG’00.
May 1997-May 1998 Research and Development
Engineer
Synopsys
Inc., Bangalore, India
Worked
on HW/SW Co-Verification tool (EAGLEI). Developed processor models for various
processors like Hitachi SH3, OakDSP. Worked on Binary Decision Diagrams for
cycle based Verilog simulator.
TEACHING
EXPERIENCE
Aug 2002-Dec 2002 Teaching Assistant
for Graduate course on Image processing (ECE447), UIUC
Gave
lectures on Fractal Image coding, Designed homeworks and machine problems,
graded them, had regular office hours.
Jan 2002-May 2002
Teaching Assistant for
Graduate/Undergraduate course on Multimedia Signal Processing (ECE371TSH), UIUC
Was
involved in designing the course including the course content, structure,
reference books, problem sets and machine problems. Gave lectures on Bayesian
networks, hidden markov models and had regular office hours.
Aug 2000 – Dec 2000 Teaching Assistant for Graduate course on Image processing
(ECE447), UIUC
Gave
lectures on hidden markov models. Designed homeworks and machine problems,
graded them, had regular office hours.
AWARDS
and HONORS
Robert T. Chien award for excellence in research in Electrical and
Computer Engineering department at UIUC for the year 2003.
Nomination
for the Best paper award in the conference – Algorithmic Learning
Theory’01 (Paper is adjudged as one of the top 3 papers of the conference).
IBM Research Fellowship for the year 2001-2002.
Presented
the system (which was build by me at Microsoft Research) as part of the Key
Note address given by Bill Gates at the International joint conference of
Artificial Intelligence’01.
Best Project of the year (1996-97) Award by IIT Delhi.
ICMI Stay Ahead Award (1996-97) by IIT Delhi for the
best project in Electrical and Computer Science department.
3rd
Rank in Entrance examination for the regional engineering colleges (out
of 200,000 applicants)
Invited
to give lectures at TsinghuaUniversity in Nov’2002.
99.8
percentile in the GATE’97 (Graduate aptitude test in Engineering)
conducted by IITs and IISc, India.
Highest
ranking in exit review while doing summer internship at Microsoft Research.
Highest
ranking in exit review while doing summer internship at Schlumberger Inc (Cairo, Egypt).
PROFESSIONAL
ACTIVITIES
Program Committee member for ICML’2003.
Program Co-chair “Very Low Bitrate Video Coding Workshop” (VLBV’98)
held at Beckman Institute, UIUC.
Student Member IEEE and member Phi Kappa Phi Honor
Society
Reviewer of IEEE Trans. on Pattern Analysis and
Machine Intelligence (PAMI) and many other International Vision conferences
including FG’2000, CVPR’2000, ICPR’2000, ICIP’2000,
ICCV’2001, ICML’2002.
INVITED
LECTURES
A brief introduction to Bayesian Networks,
Mar’2001, IIT Delhi, India.
Event Detection and Conditional Entropic HMM,
Aug’2002, Avaya Research Lab, NJ.
Event Detection and Mutual information HMM,
Aug’2002, Mitsubishi Electric Research Lab, NJ.
Introduction to Support Vector Machines,
Nov’2002, TsinghuaUniversity, Beijing, China
Activity Detection in office Environments,
Nov’2002, Institute of Automation, ChineseAcademy of Sciences, Beijing, China
PATENTS FILED
1. V. Pavlovic, Ashutosh
Garg and S.
Kasif,
“A Bayesian framework for
combining gene predictions” Patent Disclosure filed. (with HP research Lab)
2. Nuria Oliver and Ashutosh Garg,
“Maximizing Mutual Information
between observations and hidden states to minimize classification errors”
Patent Disclosure filed. (With Microsoft Research)
3. Nuria Oliver, Eric Horvitz and Ashutosh Garg,
“Layered models for Context awareness” Patent Disclosure
filed. (With Microsoft Research)
PUBLICATIONS
Journal
Papers (Refreed)
1. Vladimir Pavlovic, Ashutosh Garg and
Simon Kasif, “A Bayesian Framework for combining gene predictions,”
Journal of BioInformatics, Jan’2002
2. Ashutosh Garg, Vladimir Pavlovic, Jim Rehg, and Thomas S. Huang,
“Boosted Dynamic Bayesian Networks for multimodal Speaker Detection”
submitted to Proceedings of IEEE.
3. Gerasimos Potamianos, Chalapathy Neti, Guillaurne Gravier,
Ashutosh Garg, Andrew Senior, “Recent Advances in the Automatic
Recognition of Audio-Visual Speech” submitted to Proceedings of IEEE.
4. Ira Cohen, Nicu Sebe, Larry Chen,
Ashutosh Garg, Thomas S. Huang, “Facial Expression Recognition from
Video Sequences: Temporal and Static Modeling”, submitted to CVIU
special issue on Face recognition (currently being reviewed)
5. Ashutosh Garg, Sariel Har-Peled and Dan Roth, “Generalization
bounds and Margin distribution,” submitted to Machine Learning
Journal.
6. Ashutosh Garg, Dan Roth and Thomas S. Huang,
“Sequential models in Mulitmedia Analysis,” in preparation for
submission to IEEE Trans. on Pattern Analysis and Machine Intelligence.