# Numerical algorithms, optimisation, and computer performance: a selection of research from 2020

#### Andrea

I won’t bore you with clichés about how weird and difficult this year has been. Instead, I’d like to point out straightaway how large this year’s selection of papers is: more than 400 papers, which is more than twice the papers I selected last year.

This year, I haven’t changed my Google Scholar query:

algorithm sparse method parallel performance matrix numerical optimization graph

but I have added some more supplemental papers, as I found fit.

In the coming year, I expect to add some papers about Computational Sustainability, a field I’m getting very interested in.

## Some general trends

• This year, I found several papers about more or less “classic” mathematical optimisation applied to machine learning and deep learning. To me, this was quite predictable, but it’s always good to see how powerful and applicable some of those “classic” techniques are.
• Multi-precision algorithms are continuing to grow in popularity in the field of numerical analysis. The idea behind these algorithms is already very in deep learning.

Since we have 400 papers this year, I’m going to get some help from machine learning to analyse the data further. Click here to see the LiveScript.

From the data analysis, I see:

• among the list of papers, the topic of numerical methods applied to tensors occurred a couple of times (topic 15 in my analysis). This is most likely a new trend that I’d like to keep an eye on, even if I don’t know much about tensors;
• reading more about the topic of error analysis for multi-precision algorithms (topic 17 in my analysis) matched what I expected from last year’s post;
• the topic of matrix partitioning for parallel algorithms and distribution of matrix blocks (topic 8 in my analysis) is still occurring in my selection;
• sparsity isn’t occurring that much in the papers anymore. In practice, it’s in only in one of the topics/subtopics my analysis identified (topic 13). This might be because the papers in the selection have increased, but not the papers in this area.

## List of publications, with abstracts

This year’s papers are so many that I was unable to turn them into a WordPress-friendly list. To see the list, you can click on this link.