Deep belief network-based drug identification using near infrared spectroscopy

Abstract

Near infrared spectroscopy (NIRS) analysis technology, combined with chemometrics, can be e®ectively used in quick and nondestructive analysis of quality and category. In this paper, an e®ective drug identi¯cation method by using deep belief network (DBN) with dropout mechanism (dropout-DBN) to model NIRS is introduced, in which dropout is employed to overcome the over¯tting problem coming from the small sample. This paper tests proposed method under datasets of di®erent sizes with the example of near infrared di®use re°ectance spectroscopy of erythromycin ethylsuccinate drugs and other drugs, aluminum and nonaluminum packaged. Meanwhile, it gives experiments to compare the proposed method's performance with back propagation (BP) neural network, support vector machines (SVMs) and sparse denoising autoencoder (SDAE). The results show that for both binary classi¯cation and multi-classi¯cation, dropout mechanism can improve the classi¯cation accuracy, and dropout-DBN can achieve best classi¯cation accuracy in almost all cases. SDAE is similar to dropout-DBN in the aspects of classi¯cation accuracy and algorithm stability, which are higher than that of BP neural network and SVM methods. In terms of training time, dropout-DBN model is superior to SDAE model, but inferior to BP neural network and SVM methods. Therefore, dropout-DBN can be used as a §

Classification and Discrimination of Different Fungal Diseases of Three Infection Levels on Peaches Using Hyperspectral Reflectance Imaging Analysis

“…Although DBN showed a good performance in the classification of different fungal diseases based on spectral and imaging information, this research shows that it was not suitable for a small number of data. Similar results have been reported by Yang [ 23 ]—when the deep learning model was applied to fewer samples in each category in a multi-class classification, this led to overfitting, and conventional machine-learning technologies had certain advantages in this regard.…”

“…The program codes for the deep belief network (DBN) were written using Matlab 2017 (The Mathworks Inc., Natick, MA, USA). DBN can be efficiently trained in an unsupervised, layer-by-layer manner which comprises restricted Boltzmann machines (RBM) [ 23 ]. In this study, the DBN model had one input layer, one output layer and two hidden layers.…”

Monthly Streamflow Forecasting Using EEMD-Lasso-DBN Method Based on Multi-Scale Predictors Selection

“…Hence, one of the important steps is to calculate the corresponding parameter, , and a detailed description of the process of calculating the parameters of the DBN model can be found in [39][40][41].…”

“…where θ = w ij , a i , b j is defined as the parameter of the RBM; w ij represents the weights between visible neurons, i, and hidden neurons, j; a i is the bias parameter of ith visible neurons; is the bias parameter of ith hidden neurons; and m and n are the numbers of visible and hidden neurons, respectively. Hence, one of the important steps is to calculate the corresponding parameter" and a detailed description of the process of calculating the parameters of the DBN model can be found in [39][40][41].…”

Identification of Multi-Class Drugs Based on Near Infrared Spectroscopy and Bidirectional Generative Adversarial Networks

“…It is usually combined with chemometrics methods such as partial least squares discriminant analysis (PLS-DA) [ 3 , 4 , 5 ], linear support vector machine (Linear SVM), and other linear classifiers [ 6 , 7 , 8 , 9 , 10 ] and BP-ANN classifier [ 10 , 11 ] in a classification scenario. In recent years, some deep learning methods, such as stack sparse auto-coding (SAE) [ 12 ], deep belief network (DBN) [ 13 ], deep convolution neural network (CNN) [ 14 ], have also been reported in drug identification and classification modeling.…”