Toyota AI Seminar: Unsupervised Feature Learning via Sparse Hierarchical Representations

Machine learning has proved a powerful tool for artificial
intelligence and data mining problems. However, its success has
usually relied on having a good feature representation of the data,
and having a poor representation can severely limit the performance of
learning algorithms. These feature representations are often
hand-designed, require significant amounts of domain knowledge and
human labor, and do not generalize well to new domains. To address
these issues, there has been much interest in algorithms that learn
feature hierarchies from unlabeled data. In this talk, I will discuss
the fundamental challenges and talk about my current and future work
in developing machine learning algorithms that can learn invariant
representations from unlabeled and labeled data.
Honglak Lee is an Assistant Professor of Computer Science and
Engineering at the University of Michigan, Ann Arbor. His research
focuses on machine learning and its application across a broad range
of perception challenges, from computer vision and robotics to speech
recognition and natural language processing. His research is aimed at
developing algorithms for unsupervised and semi-supervised learning of
hierarchical features for artificial intelligence and large-scale data
mining applications. Other areas of interest include supervised
learning, probabilistic graphical models, convex optimization, and
high-dimensional data analysis. He obtained his Ph.D. in Computer
Science from Stanford University and has received ICML 2009 best
application paper award and CEAS 2005 best student paper award.