Stokastik

Machine Learning, AI and Programming

Tag: KD Tree

Fast Nearest Neighbour Search - Product Quantization

In this on-going series of fast nearest neighbor search algorithms, we are going to look at Product Quantization technique in this post. In the last post, we had looked at KD-Trees, which are effecient data structures for low dimensional embeddings and also in higher dimensions provided that the nearest neighbor search radius is small enough to prevent backtracking. Product Quantization or PQ does not create any tree indexing data structure […]

Continue Reading →

Fast Nearest Neighbour Search - KD Trees

In a few of my earlier posts "designing Q&A retrieval system" and "designing e-commerce system for similar products using deep learning", one of the primary goals was to retrieve the most similar answers (responses) and the most similar e-commerce products respectively in a fast and scalable way. The later post was pretty much generic i.e. find dense representations for each item (questions, product images etc.) using un-supervised and supervised learning […]

Continue Reading →

Designing large scale similarity models using deep learning

Finding similar texts or images is a very common problem in machine learning used extensively for search and recommendation. Although the problem is very common and has high business value to some organisations, but still this has remained one of the most challenging problems when the database size is very large such as >50GB and we do not want to lose on precision and recall much by retrieving only 'approximately' […]

Continue Reading →