Hashing Hyperplane Queries to Near Points with Applications to Large-Scale Active Learning

• Published in: IEEE transactions on pattern analysis and machine intelligence Vol. 36; no. 2; pp. 276 - 288
• Main Authors: Vijayanarasimhan, Sudheendra, Jain, Prateek, Grauman, Kristen
• Format: Journal Article
• Language: English
• Published: Los Alamitos, CA IEEE 01.02.2014; IEEE Computer Society
• Subjects: active learning; Algorithm design and analysis; Algorithmics. Computability. Computer arithmetics; Algorithms; Applied sciences; approximate nearest neighbors; Approximation algorithms; Approximation methods; Artificial Intelligence; Computer science; control theory; systems; Data Compression - methods; Data Mining - methods; Data processing. List processing. Character string processing; Databases, Factual; Euclidean distance; Exact sciences and technology; Hashing; large-scale search; Memory organisation. Data processing; Pattern Recognition, Automated - methods; Search problems; Signal Processing, Computer-Assisted; Software; Theoretical computing; Vectors; Nearest neighbour; Data analysis; Scanning; Locality; large-scale search; Hyperplane; active learning; approximate nearest neighbors; Vector space; Active system; Supervised learning; Hashing; Selection criterion; Minimal distance; Euclidean space; Large scale; Learning algorithm
• ISSN: 0162-8828; 1939-3539; 2160-9292; 1939-3539
• DOI: 10.1109/TPAMI.2013.121

We consider the problem of retrieving the database points nearest to a given hyperplane query without exhaustively scanning the entire database. For this problem, we propose two hashing-based solutions. Our first approach maps the data to 2-bit binary keys that are locality sensitive for the angle between the hyperplane normal and a database point. Our second approach embeds the data into a vector space where the euclidean norm reflects the desired distance between the original points and hyperplane query. Both use hashing to retrieve near points in sublinear time. Our first method's preprocessing stage is more efficient, while the second has stronger accuracy guarantees. We apply both to pool-based active learning: Taking the current hyperplane classifier as a query, our algorithm identifies those points (approximately) satisfying the well-known minimal distance-to-hyperplane selection criterion. We empirically demonstrate our methods' tradeoffs and show that they make it practical to perform active selection with millions of unlabeled points.