Science blogger gives some commentary on our work in unsupervised learning of materials science abstracts
Earlier this year, Tshitoyan et al. published a paper on using unsupervised machine learning of materials science journal abstracts. Based on this unsupervised learning, the algorithm used was able to identify the underlying structure of the periodic table and learn structure-property relationships. Derek Lowe, author of In the Pipeline, an independent blog from Science Translational Medicine, gives some commentary on that paper. In particular, he notes that, although “machine learning” and “artificial intelligence” have been perhaps overly hyped, the work by Tshitoyan et al. is an instance where machine learning actually does seem to be working, and although materials science is a much more orderly field than drug discovery, this work is really a step forward in terms of demonstrating success of machine learning in science. Read the full blog post here: Machine-mining the literature.