Cohorts Analysis and Visualization
Python
K-Means
K-Medoid
DBSCAN
Hypothesis test
Participated in an NIH project to build a smart trauma resuscitation decision support system.I clustered partition patients into cohorts of similar conditions and designed weight learning algorithm on case-based feedbacks then conducted statistical significance test to uncover the association between treatment patterns and patient cohorts.Paper was submitted to the analytic track of International Conference of Health Informatics 2018.
Modeling Human Affective Behavior with Deep Learning
Python
CNN
GRU
DCGAN
We applied CNN, RNN (GRU) and DCGAN models to accomplish speech-driven facial animation. We generated landmarks of the speaker from audio and video features by applying CNN and GRU and synthesized faces conditioned on landmarks by applying DCGAN.
Project based on CVPRW paper.
Fast Trajectory Replanning for Generated Mazes
C++
A trajectoty planning program extended A* algorithm to the repeated forward and backward A* algorithms which can be applied in the field of slef-driving cars. For more details, click A Project on Fast Trajectory Replanning for Computer Games
for “Introduction to Artificial Intelligence” Classes.
Sentiment Analysis
Java JavaScript HTML MapReduce
A visualization tool and a Java program implemented the MapReduce method to do Sentiment Analysis for any article.
Digits classification and Face recoginition
C++ Naive Bayes Classifier MIRA Perceptron
Trained three classifiers, Naive Bayesian Classifier, Perceptron and MIRA based on the density-detected feature extraction and got a 90% accuracy for test digit data and a 94% accuracy for face test data. For more details, click Face Detection and Digit Classification.
