What is it about?
We compared various machine learning approaches with different metrics and 8 different datasets.
Featured Image
Why is it important?
There has been lots of hype around deep learning so we wanted to see how it does versus a whole array of classic machine learning methods and we kept the fingerprint type constant using FCFP6. We performed 4 fold cross validation. Based on rank normalized scores for either metric or dataset type deep learning performed best. We also observed over training for the deep learning models.
Perspectives
This study suggested that there was still plenty of scope to explore how models perform prospectively. It should be noted that all the datasets we used had been used with Bayesian approaches and made public.
Dr Sean Ekins
Collaborations in Chemistry
Read the Original
This page is a summary of: Comparison of Deep Learning With Multiple Machine Learning Methods and Metrics Using Diverse Drug Discovery Data Sets, Molecular Pharmaceutics, November 2017, American Chemical Society (ACS),
DOI: 10.1021/acs.molpharmaceut.7b00578.
You can read the full text:
Contributors
The following have contributed to this page
