This web page allows an interface platform for the analysis of single cycle, DC voltammetric measurements using an algorithm soon to be published by the Monash electrochemistry group. The calculations are completed on the google cloud platform, with these classification algorithms using Deep Neural network to predict the most probable reaction mechanisms of the six electrochemical reaction mechanisms the models where trained on. The two models used is one (Supervised) which is trained on the simulated data with known reaction mechanisms. With the other model (Semi-supervised) is trained on data that goes through an initial unsupervised clustering algorithm to group the data based on features present before supervised training. A rough visual representation of how the deep neural network looks at the DC current can be seen in there respective feature maps outputs on the provided experimental data (shown below). Feel free to test it out on one of our provided random samples.

To submit a DC cyclic voltammogram submit a text file in the below link and press submit. If the experimental format is supported it will be passed to the machine learning models and a prediction will be made. The machine learning generally takes one minute to start up the google platform systems. After which the prediction speed will increase with the calculation then taking 20 seconds to complete. It should also be noted that when it comes the accuracy of the provided predictions the Semi-supervised model generally gives less confident predictions with a low chance of being wrong. Whereas, the Supervised gives high confidence predictions that have a tendency to return false positives.

Inputs

Prediction Table

Feature maps

Semi-supervised	Supervised

Sponsors

Website created and maintained by Luke Gundry