High-Speed Cyclic Voltammetry Regressions Using Machine Learning

The ability to rapidly regress the kinetic parameters from cyclic voltammograms is important for many laboratory automation endeavors. Inspired by how expert electrochemists can rapidly interpret cyclic voltammograms simply by their shape, here we show that convolutional neural networks (similar to those used in handwriting analysis) can successfully regress both the kinetic rate constant and transfer coefficient of cyclic voltammetry data. This type of machine learning models obtained 93.6% five fold cross validation accuracy, and could obtain both the kinetic rate constant and transfer coefficient in a few milliseconds. This is in comparison to over 3000 seconds using an optimization protocol with finite elemental analysis to regress the kinetic parameters via solving the governing differential equations. This advancement will be very useful for many electrochemical applications where high-throughput experimentation is necessary.