This is a very important point, can't be stressed enough! The effective dimensionality of data-sets varies wildly and ML models that do well on low-dimensional data might fail catastrophically in many dimensions, and vice versa ... #COMPCHEM
@ProfvLilienfeld @jchodera I take the opportunity to advertise one of our articles where we show that data-effiency also depends on the chemical diversity in the datasets. Use QM9 for your articles, but not only ;) https://t.co/x6oziLEwjg
RT @jcheminf: new: "Dataset’s chemical diversity limits the generalizability of machine learning predictions" https://t.co/kPLXJmySLL https…
RT @NakataMaho: https://t.co/fKDS7lxfuH ぼくがつくったPubChemQCデータセットを色々調べてくれてます。QM9よりよい結果がでてます。ありがたい #pubchem #pubchemqc #compchem #machinelearni…