Dihong Gong University of Florida
Data Science Research Lab
Department of Computer Science
412 Newell Drive, Room 457
Gainesville, FL 32611
Email: gongd (at) ufl (dot) edu
I am a fifth-year Ph.D. candidate in Computer Science, University of Florida.
My research interests lie in the areas of machine learning, computer vision,
information extraction, and data mining. I am currently working on multimodal knowledge
extraction from the web, advised by Dr. Daisy Zhe Wang.
My current research objective is to enable multimodal analysis for large-scale
knowledge extraction from the web. The multimodal analysis combines complementary
information clues across different modalities, which can be
beneficial for enhanced extraction accuracy. In the meanwhile, I develop scalable
parallel processing systems using Hadoop HDFS, Apache Spark, and Cuda deep learning, to enable
efficient information extraction for big data.
Multimodal Knowledge Extraction Dihong Gong, Daisy Zhe Wang (supervisor)
Primary Ph.D. research topic.
May 2015 - Present
We consider the problem of extracting instances of both text categories (e.g. persons, cities) and image categories (e.g. vehicles) from an open domain. Our goal is to design efficient and scalable algorithms to utilize text and image information and extract knowledge from the multimodal web contents. Such algorithms include: multimodal graphical models, deep learning fusion models, multimodal rules, and sparse logistic regression models based on Skip-Gram models for word-to-vec embeddings. Extensive experimental evaluations based on large-scale real-world web knowledge mining shows that our approach can achieve good improvement over the state-of-the-art algorithms. More info: Blog1, Blog2.
Computer Vision Dihong Gong, Zhifeng Li (supervisor)
Research Assistant at Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences
Aug 2011 - Dec 2011
Research topics include face recognition (cross-modality, age-invariant, video-based), and image-based human age estimation. It involves the development and application of many important machine learning concepts including: adaptive feature extraction, subspace learning, kernel methods, ensemble methods, Gaussian Process, variational inference, graphical models, deep learning, etc.
Tencent Internship (Shenzhen, China) Dihong Gong, Zhifeng Li (supervisor)
Research Scientist Intern at Tencent AI Lab
Dec 2017 - Present
Face detection and recognition, optical character recognition.
Google Internship (Mountain View, US) Dihong Gong, Na Tang (supervisor)
Tech Intern at Google Search
May 2016 - Aug 2016
Design and Implement a Parallel DocChart model to allow document annotators run in parallel within Goldmine system.
NIST NEON Datascience Evaluation Jan 2017 - Present
This project aims at developing a NIST data science contest for plant identification with Neon remote sensing data. In this project, I am responsible for defining the evaluation metrics as well as baseline algorithms for three different tasks, and coordinating implementation of baseline algorithms and development of automatic evaluation system. More info: Website, Github.
Deep Neural Networks OpenCL Implementations Dihong Gong, Siva Prasad and Siliang Xia
Work is released under BSD license.
Sep 2014 - Nov 2015
The objective of this project is to develop a device independent visual designer for
deep learning neural networks -- You design your networks with our GUI tools, and we
generate codes for you to run on a wide range of devices including GPU and CPU from
different vendors (e.g. Intel, AMD and Nvidia). To allow device-independent implementation,
we write the system with C++ and OpenCL languages. Our system is designed to support
a wide range of deep learning algorithms such as convolutional neural networks, recurrent neural networks,
deep Boltzmann machines, etc. With scalability in mind, our codes are carefully designed for optimized performance.
NIST Pre-Pilot Datascience Evaluation Dihong Gong, Daisy Zhe Wang
Sep 2015 - Jan 2016
We participated in the Pre-pilot data science evaluation organized by the National Institute of Standards and Technology (NIST), the primary goal of which is to develop and exercise the evaluation process in the context of data science. The evaluation consists of four tasks including data cleaning, data alignment, forecasting and prediction. Our DSR lab has participated the data cleaning and traffic event prediction tasks, and submitted several running systems (most of which are based on student projects in our data science class) of different algorithms and configurations. More info: Blog1.
Biological Networks Learning Dihong Gong, Ahmet Ay, Tamer Kahveci (supervisor)
Research Assistant at Bioinformatics Lab, University of Florida
Mar 2014 - Sep 2014
We proposed network-based classification model for cancer prediction using gene expression, and compare our model to other classification models such as support vector machines, naive bayes classifier, k nearest neighbors, C4.5 decision trees, random forest, etc. Additionally, we also proposed an efficient algorithm (D-Hiden) to find the hierarchy of the genes in dynamically evolving gene regulatory network topologies. I was reponsible for the implementation of all the algorithms, and conducting comparison evaluation.
TTCN-3 Compiler Design and Implementation Dihong Gong, Guobin Chen, Sihai Zhang (supervisor)
Undergraduate student at University of Science and Technology of China
Oct 2009 - Feb 2011
We designed an efficient compiler implementation framework for TTCN-3 programming language, and implemented TTCN-3 compiler that compiled TTCN-3 source codes into ASM codes. This work is from National innovation experiment program for college students of China, supported by National Ministry of Education of China with Grant 091035837.
"Multimodal Learning for Web Information Extraction",
D. Gong, D. Z. Wang, Y. Peng
ACM Multimedia (Full Research Paper), 2017.
Rank A+ conference!
"Multi-feature Anisotropic Orthogonal Gaussian Process for Automatic Age Estimation",
Z. Li, D. Gong, K. Zhu, D. Tao, X. Li
ACM Transactions on Intelligent Systems and Technology, 2017.
(Impact Factor: 3.19)