|
|
Joseph F. JaJa Department of Electrical and Computer Engineering,
Institute for Advanced Computer Studies, University of Maryland, College Park |
Research Areas: Machine Learning with applications to biomedical data, Data Science,
Computational Neuroscience, Parallel Computing, Visualization.
Biographical
Sketch:
Professor: Electrical and Computer
Engineering, and Institute for Advanced Computer Studies, University of
Maryland.
2018 – 2022: Interim Chair,
Department of Electrical and Computer Engineering, University of Maryland,
College Park.
2010-2011: Interim VP and CIO,
University of Maryland, College Park.
1994-2004: Director of the University
of Maryland Institute for Advanced Computer Studies (UMIACS)
2002-2004: Interim Director of the
Center for Bioinformatics and Computational Biology
Ph.D. 1977, M.S. 1976, Applied
Mathematics, Division of Engineering and Applied Physics, Harvard University.
Over 250 refereed publications in
parallel and distributed computing, data intensive computing, combinatorial
algorithms, algebraic complexity, and VLSI architectures.
IEEE and ACM Fellow
Current
Projects:
Collaborative Research between IHC and FDA. This collaboration involves two main projects focusing on
out of distribution detection for AI-enabled medical devices, and learning with
limited data. Initial exploration has involved mammography images (especially,
the Emory Breast Cancer Mammogram Dataset EMBED), Chest X-Ray images, and skin
legion images.
Machine Learning and Data Science. This research involves a number of efforts related to
adversarial machine learning, disentangled representations, federated learning,
lifelong learning, and techniques related to computational neuroscience. The
latter project addresses the structural and functional organization of the
brain and the characteristics of the corresponding brain networks using DTI and
fMRI data.
Parallel Computing.
Development of parallel algorithms and their implementations on current and
emerging heterogeneous multicore/GPU platforms. Applications of interest range
from scientific computing to large scale graph –theoretic problems to
machine learning algorithms for big data.
Partnership with the Laboratory for Telecommunication
Sciences.
This partnership involves collaborative efforts involving a broad set of
research areas and outreach activities. Current research projects cover topics
in machine learning, quantum networks, wireless, network security, and
visualization.
Teaching:
Spring 2024: ENEE 436: Foundations
of Machine Learning
Fall 2023: ENEE 290: Introduction to
Ordinary Differential Equations and Linear Algebra
Fall 2022: ENEE759G: Unsupervised
Learning
Fall 2020 and 2021: ENEE 436:
Foundations of Machine Learning
Fall 2019: ENEE759G: Unsupervised
Learning
Fall 2018, Spring
2019: ENEE101: Introduction to Electrical and Computer Engineering
Selected
Recent Publications:
HoloCamera: Advanced
Volumetric Capture for Cinematic-Quality VR Applications, J. Heagerty, S. Li, et. al., accepted
to IEEE Transactions on Visualization and
Computer Graphics, 2024.
ProtoVAE: Prototypical
Networks for Unsupervised Disentanglement, V. Patil, M. Evanusa,
and J. JaJa, arXiv.2305.09092 https://arxiv.org/abs/2305.09092,
2023.
Improving Graph
Neural Network with Learnable Permutation Pooling, Y. Jin and J.
JaJa, 2022 IEEE Conference on Data Mining Workshops, Nov. 28-Dec 1, 2022,
Orlando, FL.
TAG: Boosting
Text-VQA via Text-aware Visual Question-answer Generation, J. Wang, M.
Gao, Y. Hu, F. Selvaraju, C. Ramaiah,
R. Xu, J. JaJa, and L. Davis , to appear in the
Proceedings of BMVC 2022, Nov. 21-24, London, UK.
DOT-VAE:
Disentangling One Factor at a Time using Variational
Autoencoders,
V. Patil, M. Evanusa, and J. JaJa, to appear in the
Proceedings of the 2022 ICANN conference, Bristol, UK 6-9 September, 2022.
FedNet2Net:
Saving Communication and Computations in Federated Learning with Model Growing, A. Kundu and J.
JaJa, Proceedings of the 2022 ICANN, Bristol, UK, 6-9 September, 2022.
Class-Similarity
Based Label Smoothing for Confidence Calibration, C. Liu and J. JaJa, Proceedings of the 30the International Conference
on Artificial Neural Networks (ICANN’2021), 2021.
Learning Brain Dynamics for Decoding and
Predicting Individual Differences, J. Misra,
S. Surampudi, M. Venkatesh, C. Limbachia,
J. JaJa, and L. Pessoa, accepted to PLOS
Computational Biology, 2021.
Graph
Coarsening with Preserved Spectral Properties, Y. Jin, A. Loukas, and
J. JaJa, accepted to The 23rd
International Conference on Artificial Intelligence and Statistics (AISTATS
2020).
Comparing
Functional Connectivity Matrices: A Geometry-Aware Approach applied to
Participant Identification, M. Venkatesh, J. JaJa, and L. Pessoa, Neuroimage, vol 207, Feb 2020.
Towards
a Physiological Scale of Vocal Fold Agent-based Models of Surgical Inquiry and
Repair: Sensitivity Analysis, Calibration and Validation, A. Garg, S. Yuen, N.
Seekhao, G. Yu, J. Karwowski, M. Powell, J. Sakata,
L. Mongeau, J. JaJa, and N. Li-Jessen, Applied
Sciences, 9(15), 2019.
Geodesic
Distances between Functional Connectivity Matrices: a Geometry-aware Approach,
M. Venkatesh, J. JaJa, and L. Pessoa, Poster Presentation, Neuroscience 2019, October 19-23, Chicago, IL.
Analysis and Forecasting for Traffic Flow
Data, Y. Wang and J. JaJa, Sensors
and Materials, 31(6), 2143-2154, 2019.
Feature Prioritization and Regularization
Improve Standard Accuracy and Adversarial Robustness, C. Liu and J.
JaJa, to appear in Proceedings for the
International Joint Conference on Artificial Intelligence (IJCAI), August
2019.
Brain Dynamics and Temporal Trajectories
during Task and Naturalistic Processing, M. Venkatesh, J. JaJa, and L.
Pessoa, Neuroimage, 2019.
High-Performance Agent-based Modeling
Applied to Vocal Fold Inflammation and Repair, N. Seekhao, C. Shung, J.
JaJa, L. Mongeau, and N. Li-Jessen, Frontiers in Physiology, April 2018, Vol.
9.
LEICA: Laplacian Eigenmaps for Group ICA
Decomposition, C. Liu, J. JaJa, and L. Pessoa, Neuroimage, 2017.
In Situ Visualization for 3D Agent-Based
Vocal Fold Inflammation and Repair Simulation, N. Skeekhao,
J. JaJa, L. Mongeau, and N. Li-Jessen, accepted to Supercomputing Frontiers and Innovations, 2017.
Real-time Agent-Based Modeling and Simualtion with in-situ Visualization of Complex Biological
Systems: A Case Study on Vocal Fold Inflammation and Healing, N.
Seekhao, C. Shung, J. JaJa, L. Mongeau, and N. Li, Proceedings of the HiCOMB 2016 Workshop,
May 2016, Chicago.
A High Performance Implementation of
Spectral Clustering on CPU-GPU Platforms, Y. Jin and J. JaJa, Proceedings of the Parallel Computing and
Optimization (PCO 2016) Workshop, May 2016, Chicago.
Linking “Toxic Outliers” to
Environmental Justice Communities, M. Collins, I. Munoz, and J. JaJa, Environmental Research Letters, December
2016. Won the ERL Best Article for 2016.
Connectivity-Based
Brain Parecellations: A Connectivity-Based Atlas for Schizophrenia Research,
Q. Wang, R. Chen, J. JaJa, Y. Jin, E. Hong, and E. Herskovits, Neuroinformatics, October 2015.
A
Data-Driven Approach to Extract Connectivity Structures from Diffusion Tensor
Imaging Data, Y. Jin, J. JaJa, R. Chen, and E. Herskovits, Proceedings of the 2015 IEEE International
Conference on Big Data (IEEE BigData 2015), Oct 29 – Nov 1, 2015, Santa Clara,
CA.
Achieving Native GPU Performance for
Out-of-Card Large Matrix Multiplication, J. Wu and J. JaJa, Parallel Processing Letters, 2016.
Resting State Dynamic Functional
Network Analysis in Mild Traumatic Brain Injury, W. Hou, C. Chandler, J. JaJa,
and R. Gullapalli, International
Society for Magnetic Resonance in Medicine ISMRN 2015, Toronto, Ontario, May 30—31, 2015.
Optimized FFT Computations on Heterogeneous
Platforms with Application to the Poisson Equation. J. Wu and J. JaJa, Journal of Parallel and Distributed
Computing, 2014.
From Maxout to
Channel-Out: Encoding Information on Sparse Pathways, Q. Wang and J.
JaJa, Proceedings of the International
Conference on Artificial Neural Networks, 15-19 September, 2014, Hamburg,
Germany.
High Performance FFT Based Poisson Solver
on a CPU-GPU Heterogeneous Platform, J. Wu and J. JaJa, International Parallel and Distributed
Processing Symposium (IPDPS), May 2013, Boston, Massachusetts.
Hierarchical Exploration of Volumes Using
Multilevel Segmentation of the Intensity-Gradient Histograms, C. Yiu Ip,
A. Varshney, and J. JaJa, VIS 2012, October 2012 (won Best Paper Award for SciVis 2012)
An
Effective Approach to Temporally Anchored Information Retrieval, Z. Wei
and J. JaJa, UMIACS-TR-2012-10, University of Maryland, College Park, August,
2012.
Optimized Strategies for Mapping
Three-dimensional FFTs onto CUDA GPUs, J. Wu and J. JaJa, Proceedings of Innovative Parallel Computing (INPAR) Workshop, San Jose, CA, May 13-14, 2012.
Constructing Inverted Files: To MapReduce or Not Revisited, Z. Wei and J. JaJa, Technical Report, University of
Maryland, College Park, 2011. Also, to appear in IEEE Transactions on Parallel and Distributed Computing.
A
Fast Algorithm for Constructing Inverted Files on Heterogeneous Platforms,
Z. Wei and J. JaJa, International
Parallel and Distributed Processing Symposium (IPDPS), May 2011, Anchorage,
Alaska.
Optimization
of Linked List Prefix Computations on Multithreaded GPUs Using CUDA, Z
Wei and J. JaJa, International Parallel and Distributed Processing Symposium
(IPDPS), April 2010, Atlanta, GA.
Techniques
to Audit and Certify the Long Term Integrity of Digital Archives, S.
Song and J. JaJa, International Journal of Digital Libraries, 2010.
--------------------------------------------------------------------------------------------------------------
Last Updated, June, 2022.