Estimation of pure component parameters of PC-SAFT EoS by an artificial neural network based on a group contribution method

Multitask Neural Network for Mapping the Glass Transition and Melting Temperature Space of Homo- and Co-Polyhydroxyalkanoates Using σProfiles Molecular Inputs

“…Increasing the number of layers from one to two generally improves the prediction accuracy. , Consequently, testing the addition of a second hidden layer for predicting T g and T m of PHA polymers is essential for the model’s success. Therefore, the total number of neurons in both layers was varied between 10 and 50, with a minimum of 5-5 and a maximum of 25-25.…”

“…86,87 Increasing the number of layers provides a shortcut to increasing the capacity of the model with fewer resources; generally, two hidden layers are sufficient for most problems pertaining to a single output response. 88 A minimal number of neurons may lead to an under-fitted model, which will have limited training and testing data accuracy. On the other hand, a model with a high number of neurons will be over-fitted, resulting in a high training accuracy with a low testing performance.…”

“…For instance, previous studies used either one or two hidden layers , for predicting the properties of deep eutectic solvents. Thus, the most prevalent approach in the literature is trial and error . For this reason, various network architectures with one and two hidden layers and varying numbers of neurons were studied to propose an effective multitask-ANN model in predicting the T g and T m of PHAs.…”

“…Thus, the most prevalent approach in the literature is trial and error. 88 For this reason, various network architectures with one and two hidden layers and varying numbers of neurons were studied to propose an effective multitask-ANN model in predicting the T g and T m of PHAs. The hidden neurons in hidden layers 1 and 2 (H n, l and HH n, l ) are expressed by eqs 3 and 4, respectively, where W m, n, l represents the weight coefficient between each input neuron (m) and the hidden neuron (n), l represents the hidden layer, and b represents the bias of the neuron.…”

Machine-learning-assisted molecular design of phenylnaphthylamine-type antioxidants

“…S3 (ESI †). [62][63][64][65] The LevenbergÀMarquardt method was used in the optimization algorithm with the learning rate set to 0.01. In the ANN optimization process, the mean square error (MSE) was employed as the loss function; it is described by eqn (1): 66…”