The basis for a novel, and quite general, similarity-based approach for representing chemical spaces is presented. A key feature of the method, which is derived from generalized Fourier series, is an ansatz that equates inner products between 'molecular basis vectors' to the similarities of the corresponding molecules. Importantly, there is no need to know the actual form of the molecular basis vectors since one computes the inner product or Gramian matrix directly from the molecular similarities of the basis molecules, which is quite analogous to kernel-based learning methods where construction of the Gramian matrix does not always require a detailed knowledge of the feature space.

Molecular basis sets: A general similarity-based approach for representing chemical spaces

239th National Meeting (2010)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.