Charting ML Publications in Science

As machine learning has grown in popularity and domain of applicability, charts showing the growth have popped up in review articles, presentations, and more. Many of these charts seem to have different numbers (though showing the same trend!), and are often without provenance. Thus, this repository was made to act as a resource for those looking to dig into the numbers further, to quickly customize our existing plots, or to simply find a ready-made chart for their own usage.

2023 Results

Table showing the count and Compound Annual Growth Rate (CAGR) on the raw number of articles published over different periods for each domain. CAGR-N represents the CAGR over N years.

Domain	year	count	CAGR-1 (%)	CAGR-2 (%)	CAGR-5 (%)
Materials Science	2023	7471	18.5	31.5	66.7
Chemistry	2023	10134	13.9	22.6	37.4
Physics	2023	8023	15.6	24.4	55.9

The Charts

Note that the charts come in two forms, normalized and count. Normalized plots take into account the relative size of the domain of interest (i.e., the number of matching articles / the total number of articles in a domain).

Normalized Line Plot

Raw Count Line Plot

The Domains

• Materials Science
• Chemistry
• Physics

Methodology

• Web of Science topic matching and matching of domain. Exact queries are provided in the data directory.

ID	Service	Query
1	Web of Science	TS=("machine learning" OR "informatics" OR "deep learning" OR "cheminformatics" OR "artificial intelligence" OR "chemoinformatics" OR "QSAR" OR "QSPR") AND WC="{Domain}"

Customization

• All routines to create the plots are provided in the repo notebooks.
• Customize the figure closest to your heart, and it will be saved out to the output directory.
• Don't like the queries used? Simply output your data to the same format used in /data/1.csv and re-use the plotting tools.

TODO (Pull requests welcome)

• Automate data pull through Web of Science API
• Consider a Google Scholar implementation
• Improve plot consistency and styling
• Add statistics for other domains
• Consider adding other common plots
• Fix plots now that there are more domains in the data

Cite

Ben Blaiszik, “2021 AI/ML Publication Statistics and Charts”. Zenodo, Sep. 07, 2022. doi: 10.5281/zenodo.7057437.

@software{ben_blaiszik_2023_7713954,
  author       = {Ben Blaiszik},
  title        = {{blaiszik/ml\_publication\_charts: AI/ML Publication 
                   Statistics for 2022}},
  month        = mar,
  year         = 2023,
  publisher    = {Zenodo},
  version      = {2023.03},
  doi          = {10.5281/zenodo.7713954},
  url          = {https://doi.org/10.5281/zenodo.7713954}
}