Figure 3
From: Learning physical properties of liquid crystals with deep convolutional neural networks
Predicting the pitch length of cholesteric liquid crystals with convolutional neural networks. (A) Illustration of the network architecture used for classifying the pitch values. This network has the same general structure used for classifying phases and predicting the order parameter. The difference is in the last layer that is now composed of 8 nodes with softmax activation functions. (B) Training and validation scores (accuracy, the fraction of correct classifications) as a function of the number of training epochs. We use 15% of data as test set and the remaining is divided into training (80%) and validation (20%) sets (all obtained in a stratified manner). (C) Confusion matrix obtained by applying the trained network to the test set. The diagonal form shows that the trained network achieves a perfect classification of all pitch values in the test set. (D) Accuracy estimated from the test set as a function of the number of convolution (and max-pooling) blocks nb in the architecture (panel A corresponds to nb = 4). The markers represent the average values obtained from ten realizations of the training procedures, and the error bars are 95% confidence intervals.
