You are here

BigData: Probabilistic Methods for Efficient Search and Statistical Learning in Extremely High-Dimensional Data

Monday, October 8, 2012 -
4:00pm to 4:50pm
KEC 1001

Speaker Information

Ping Li
Assistant Professor
Department of Statistical Science
Cornell University


<p>This talk will present a series of work on probabilistic hashing methods which typically transform a challenging (or infeasible) massive data computational problem into a probability and statistical estimation problem. For example, fitting a logistic regression (or <span data-scayt_word="SVM" data-scaytid="3">SVM</span>) model on a dataset with billion observations and billion (or billion square) variables would be<br>difficult. Searching for similar documents (or images) in a repository of billion web pages (or images) is another challenging example. In certain important applications in the search industry, a web page is often represented as a binary (0/1) vector in billion square (2 to power 64) dimensions. For those data, both data reduction (i.e., reducing number of nonzero entries) and dimensionality reduction are crucial for achieving efficient search and statistical learning.</p><p>This talk will present two closely related probabilistic methods: (1) b-bit <span data-scayt_word="minwise" data-scaytid="5">minwise</span> hashing and (2) one permutation hashing, which simultaneously perform effective data reduction and dimensionality reduction on massive, high-dimensional, binary data. For example, training an <span data-scayt_word="SVM" data-scaytid="4">SVM</span> for classification on a text dataset of size <span data-scayt_word="24GB" data-scaytid="7">24GB</span> took only 3 seconds after reducing the dataset to merely <span data-scayt_word="70MB" data-scaytid="8">70MB</span> using our probabilistic methods. Experiments on close to <span data-scayt_word="1TB" data-scaytid="9">1TB</span> data will also be presented. Several challenging probability problems still remain open.<br><br>Key references:<br>[1] P. Li, A. Owen, C-H Zhang, On Permutation Hashing, NIPS 2012;<br>[2] P. Li, C. <span data-scayt_word="Konig" data-scaytid="10">Konig</span>, Theory and Applications of <span data-scayt_word="b-Bit" data-scaytid="11">b-Bit</span> <span data-scayt_word="Minwise" data-scaytid="12">Minwise</span> Hashing, Research Highlights in Communications of the ACM 2011.</p>

Speaker Bio

Ping Li is an Assistant Professor in the Department of Statistical Science at Cornell University. His research interests include BigData, randomized algorithms, boosting and trees, information retrieval, etc. Ping Li won a prize in the Yahoo! 2010 Learning to Rank Grand Challenge. He is also a recipient of the ONR (Office of Naval Research) Young Investigator Award in 2009.