[Yoshitaka Yamamoto] Resource-oriented Approach for Extracting Deep Knowledge from Big Data Streams

Presto Researcher

Yoshitaka Yamamoto
Graduate School Department of Interdisciplinary Research , University of Yamanashi
Assistant Professor

Outline

This work studies online techniques for mining deep knowledge from high-dimensional data streams with numerical transactions. Deep knowledge denotes latent and high-ordered relations, rules or patterns that frequently appear in data streams. It partially corresponds to high-level features, which has been extensively studied in the context of image or sound data processing. Those research interests are focused on deep learning from unlabeled data set along with its recent success.
Although deep learners can identify high-level objects, it may be difficult to provide any explanation to them. In this work, we aim at formalizing and extracting deep knowledge in the context of online data mining. We develop new data management framework for handling high-level objects as formalized knowledge, and then integrate it with novel online mining techniques to detect latent and high-ordered relations, rules or patterns hidden in data streams.

Quick Access

Quick Access

 News

arrow On-going

arrow Completed

Program

  • CREST
  • PRESTO
  • ACT-I
  • ERATO
  • ACT-C
  • ACCEL
  • ALCA
  • RISTEX
Finish programs
  • Pamphlet
  • ProjectDB
  • GlobalActivity
  • Diversity-EN
  • OS_Policy-EN
  • Question-E