Glossary term
Glossary term
Evaluation and Benchmarks
The number of elements set to zero (or null) in a vector or matrix divided by the total number of entries in that vector or matrix. For example, consider a 100-element matrix in which 98 cells contain zero. The calculation of sparsity is as follows:
Feature sparsity refers to the sparsity of a feature vector; model sparsity refers to the sparsity of the model weights.
Created for this library
A risk modeling team uses L1 regularization to encourage sparsity in its scorecard coefficients for interpretability.
A research team studies sparsity patterns in attention to design efficient transformer variants.
An ML team prunes weights to introduce sparsity in its production model so it can run faster on supported hardware.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License