A curious correspondence between sparse and low-rank matrices and its myriad practical uses

Add to your list(s) Download to your calendar using vCal

• Lawrence Saul (None / Other)
• Thursday 24 July 2025, 09:30-10:30
• External.

If you have a question about this talk, please contact nobody.

OGGW05 - Geometric and combinatorial methods in the foundations of computer science and artificial intelligence

We consider when a sparse nonnegative matrix can be recovered, via a simple elementwise nonlinearity, from a real-valued matrix of significantly lower rank. We show that this question arises naturally in many problems of high dimensional data analysis, and for a particular choice of nonlinearity, we describe an algorithm, known as subzero matrix completion, to discover these low-rank representations. As illustrative examples, we use the algorithm to analyze the synaptic weight matrix of the fruit-fly connectome and the co-occurence statistics of words in natural language. Finally, we discuss the challenges of scaling this algorithm to very large matrices, as well as recent progress on overcoming these challenges.

This talk is part of the Isaac Newton Institute Seminar Series series.

This talk is included in these lists:

• All CMS events
• External
• Featured lists
• INI info aggregator
• Isaac Newton Institute Seminar Series
• School of Physical Sciences
• bld31

Note that ex-directory lists are not shown.