# Low Rank Matrix Factorization
Take a look at [this section](https://pytorch.org/blog/inside-the-matrix/#2d-summed-outer-products) of the referenced blog post. The key idea is this sentence:
> Among other things this builds intuition for why “low-rank factorization” - i.e. approximating a matrix by constructing a matmul whose arguments are small in the depth dimension - works best when the matrix being approximated is low rank
---
Date: 20230927
Links to:
Tags:
References:
* [Inside the Matrix: Visualizing Matrix Multiplication, Attention and Beyond | PyTorch](https://pytorch.org/blog/inside-the-matrix/)