Quoc V. Le

Quick links to:
All publications (Google Scholar version) -- Almost complete bibfile -- Coauthors -- Professional services -- Cool demos

Infomation:

PhD Student, AI Lab, Computer Science Department, Stanford University.
Advisor: Professor Andrew Ng.
Address: Room 110A, Gates Building, Stanford CA 94305.
Email: someone@somewhere where someone is quocle and somewhere is stanford.edu
I did my undergraduate at ANU & NICTA (Canberra, Australia), under the supervision of Professor Alex Smola.
I was also a research visitor at Dept Schölkopf, Max Planck Institute for Biological Cybernetics (Tübingen, Germany).

• Q.V. Le, A. Karpenko, J. Ngiam, A.Y. Ng.
  ICA with Reconstruction Cost for Efficient
  Overcomplete Feature Learning.
  NIPS, 2011. [PDF], [Appendix], [Code],
  Topics: neural networks, unsupervised feature learning, ICA, computer vision

• Q.V. Le, J. Ngiam, A. Coates, A. Lahiri, B. Prochnow, A.Y. Ng.
  On optimization methods for deep learning.
  ICML, 2011. [PDF], [Supplementary document], [More info]
  Topics: neural networks, optimization

• Q.V. Le, W.Y. Zou, S.Y. Yeung, A.Y. Ng.
  Learning hierarchical spatio-temporal features for action recognition
  with independent subspace analysis.
  (Old title: Stacked Convolutional Independent Subspace Analysis for Action Recognition.)
  CVPR, 2011. [PDF] (Oral presentation [Slides],[Talk Video])
  [Appendix], [Code], [Notes], [More Info].
  The code can let you train features on your unlabelled dataset.
  But if you do not want to train your features, you can
  download features on already trained on Hollywood2 [Features]
  Topics: neural networks, unsupervised learning, action recognition, invariances

• Q.V. Le, J. Ngiam, Z. Chen, D. Chia, P. Koh, A.Y. Ng
  Tiled Convolutional Neural Networks. NIPS, 2010. [PDF]
  [TCNN code], [Invariance visualization].
  NIPS poster [PPT], [PNG]
  Topics: neural networks, unsupervised learning, Topographic ICA, object recognition.

• Code for training deep autoencoder with L-BFGS (See this paper; this implementation is not optimized for speed)
• Stacked ISA for Videos (state-of-the-art video features, see our CVPR 2011 paper)
• TCNN code (improving convolutional neural networks with untied weights).
• Invariance visualization (visualization techniques for optimal stimuli and inviariances).
• Proximal optimization (stabilizing online gradient descent, such as Pegasos, with proximity).
• Some code is in ELEFANT or also in MLOSS (a general machine learning toolbox for kernel methods, graphical models, decision trees etc.)
• Some other is in BMRM (probably one of the first Map-Reduce style machine learning toolboxes).
• CoFiRank (a large scale collaborative ranking toolbox).
• I previously contributed to ROS (Robot Operating System).

• Me at Kioloa.
• STAIR project