Multimedia Big Data Mining and Deep Learning

The Data mining, Database & Multimedia Research Group, University of Miami

Research Assistant

August 2013 – present

Research areas: multimedia big data mining, deep neutral networks, imbalanced data analysis

Published 10+ high quality papers on big data analysis and concept retrieval

  • Integrated a bootstrapping approach for imbalanced data classification problem in neutral networks
  • Proposed an idea to feed low-level features to CNNs for saving training time
  • Proposed a scalable classifier ensemble framework assisted by a set of judgers to integrate the outputs from multiple classifiers for multimedia big data classification on Spark
  • Explored and utilized inter-concept correlations among concepts for information retrieval
  • Engaged in several different directions in big data analysis and proposed some useful ideas to help other researchers and enhance the performance of neutral networks

Built an experimental big data analysis environment

  • Installed Hadoop on one master node and two slave nodes, with two works on each node
  • Installed a NoSQL database, HBase and stored multimedia big datasets like images
  • Deployed Spark on the environment, which can run in standalone or yarn model
  • Used Caffe to exact features in different layers, tune existing models, and classify images
  • Used Theano to train different kinds of neutral networks with different configurations
  • Deployed DL4J on Spark to effectively train neutral networks by multiple workers
  • Tested most popular deep learning tools and fed them with different types of data

Developing a website for concept detection and information retrieval using big data techniques

Multimedia Processing Lab, National Taiwan Ocean University

Research Assistant

September 2011 – August 2013

Research interests: Image Processing and Computer Vision