Research

Currently, my research focus on the application of data science, statistics and machine learning methods in solving real-world problems. Below are the details of my research experience.

Research Experience

• Effect of Fumigation on Microbiomes in Soil , Jan 2023 - Present
  • This is my first research project. The project is about analyzing the effect of a toxic fumigant (Methyl Bromide) on the microbiome communities in the soil through various statistical inference and microbiome network inference techniques. The dataset is a high-dimensional, sparse and large dataset. The goal is to identify if there is change in microbiome communities caused by the fumigation. <!– * The following analysis were performed
  • Data Visualization using Non-metric Multidimensional Scaling (NMDS) and Heatmap with agglomerative clustering to identify the patterns between fumigated and non-fumigated samples.
  • Using permutational multivariate analysis of variance (PERMANOVA) to test the hypothesis if there is difference between fumigated and non-fumigated samples.
  • Constructed linear model to identify the relationship between alpha diversity (Shannon index and Inverse Simpson Index) with several predictors (fumigation, treatment, etc).
  • Build microbiome networks using Glasso, Correlation and Dissimilarity matrix, CARlasso algorithm and conducted network inference to identify the interactions between taxas and with predictors. –>
  • Advisor: Claudia Solís-Lemus, PhD student Evan Gorstein.
  • Github Link
  • Paper and further analysis in progress
• Wildlife Sound Identification, July 2023 - Present
  • The project is about identifying wildlife species based on their sound. The sound data is collected by soundscape in tropical rainforest. The goal is to develop a pipeline to be able to identify where the sound is, and to be able to identify the species.
  • Developed a pipeline to process the sound data and their labels (bounding boxes and classes).
  • Currently developing a Faster R-CNN model to identify the bounding boxes of the animal sound from the spectrogram.
  • Advisor: Claudia Solís-Lemus.
  • Github Link (Most work is still in progress)
• Deep Learning for Medical Image Classification , Feb 2023 - May 2023
  • This is my second research project. The project is about identify aggressive tumor features in Renal cell carcinoma using learning-based CV techniques. The previous group has extracted image slices with most learnable features (non-zeros on the image matrix). My job is to build and imporve a deep CNN to classify the images (high risk or low risk).
  • Developed, and optimized a convolutional neural network using PyTorch for feature extraction from images, improved test accuracy to ~90%.
  • Evaluated the CNN performance using various techniques including compare against a random model, permutation tests, evaluate the performance on different datasets, confusion matrix, etc.
  • Advisor: Dane Morgan, collaborated with Yuqing Wang and Ankit Joju
  • Full Project Description: link (You might need to scroll down to see the project description for this specific project I am working on.)