  |
Deep
Graph Learning
- Develop
efficient deep learning algorithms for processing graph data
- Develop
novel generative models for graph data generation with sematic
guide
- Novel
solutions for social data analysis, molecular informatics,
complex image-omics data analysis
|
 |
Machine
Learning for Drug Discovery
- Develop
efficient machinelearning algorithms for molecule data
- Develop
novel generative models for molecue graph generation
- Protein
folding, TCR-pMHC binding, Antibody design and optimization
- Novel
algorithms for chemical sysnthesis and retrosysthesis
|
 |
Deep
Learning for Survival Prediction
- Develop
novel nonlinear methods for censored regression problems
- Deep
Multistance Learning for survival prediction from big pathological
Images
- Novel
learning methods for small sample problems
- Deep
learning methods for multimodal censored data
|
 |
Big
Image-Omics Data Analytics
- Connections
between morphology and prognosis
- Big
pathological Image analytics for clinical outcome prediction
- High-dimensional
molecular profiling data analytics
- Integrating
pathological image data with molecular profiling data
- Generating
image or omics data from each other
|
|
Human/Facial
Behavior
- Real-time
face tracking with a web camera or Kinect
- Fatigue
detection by tracking slow eyelid closure and blinking
- Dyadic
Synchrony as a Measure of Trust and Veracity
-
Facial expression recognition
|
  |
Structure
Sparsity: Theorems, Algorithms and Applications
- Structured
sparsity theorems give the insight when known structure/sparse
priors
- Convex
and greedy algorithms for strcutured sparsity recovery problem
- Sucessful
applications on Compressive sensing on graph structured sparse
data
- Sucessful
applications on video forground detection and abnormal detection
|
  |
Magnetic
Resonance Imaging
-
Compressive Sensing techniques for accelerated MR imaging
- Fast
Image reconstruction method for compressive sensing MRI
- Gradient
sparse and wavelet sparse prior for MRI
|
 |
3D/4D
Modeling, Simulation and Segmentation
- Deformal
moded for cardiac/lung/liver/brain segmentation
- 4D
high resolution cardiac images acquired by the 320 multi-detector
CT
- 4D
cardiac reconstruction of the surface of the left ventricle
(LV) for a full cardiac cycle.
- Enabling
to investigate functional significance in health and disease
|
 |
Non-Intrusive
Load Monitoring (NILM)
- How
to fully exploit the inherent characteristics of each appliance
in a specific functional mode?
- How
to derive efficient algorithms for low-sample-rate data of
each appliance to enhance model scalability for Low-frequency
Energy Disaggregation?
- How
to investigate effiecient inference algorithms to learn the
latent states from measured aggregation data?
|