جستجوی مقالات مرتبط با کلیدواژه « activation function » در نشریات گروه « صنایع غذایی »

Today, plant and commercial gums are used to improve the rheological, textural and sensorial properties of food. Basil seeds have significant amounts of gum (mucilage) with good functional properties that after extracting from the seeds and drying, can be used as a powder in the formulation of various products. In this study, to drying of basil seed mucilage, infrared radiation (IR) method was used. The effect of infrared lamp power (150, 250 and 375 W), the distance of sample from lamp (4, 8 and 12 cm) and mucilage thickness (0.5, 1 and 1.5 cm) on drying time of basil seed mucilage were investigated. The results of basil seed mucilage drying using infrared method showed that with increasing lamp power and decreases in sample distance from the heat source, drying time was decreased. With increasing in the lamps distance from 4 to 12 cm, the average drying time of basil seed mucilage increased from 131.37 minutes to 336.41 minutes. With increasing sample thickness from 0.5 to 1.5 cm, the average drying time of basil seed mucilage increased from 103.67 to 367.67 minutes. The process was modeled by an artificial neural network with 3 inputs (lamp power, lamp distance and thickness) and 1 output (drying time). The results of artificial neural network modeling showed that a network with 8 neurons in a hidden layer and using the sigmoid activation function could predict the drying time of basil seed mucilage using the infrared dryer (r=0.96). The results of sensitivity analysis by optimal neural network showed that sample thickness is the most effective factor in controlling the drying time of basil seed mucilage.

This research was aimed to study the effects of loading force and storage period on the physiological characteristic of pears. In this experiment, the pears were subjected to quasi-static loading (wide-edge and thin-edge) and different storage periods (5, 10 and 15 days). The amounts of the fruits’ total phenol, antioxidant and vitamin C contents were evaluated after each storage period. In the present study, multilayer perceptron (MLP) artificial neural network featuring a hidden layer and two activating functions (hyperbolic tangent-sigmoid) and a total number of 5 and 10 neurons in each layer were selected for the loading force and storage period so that the amounts of the total phenol, antioxidants and vitamin C contents of the fruits could be forecasted. According to the obtained results, the highest R2 rates for thin-edge and wide-edge loading in a network with 10 neurons in the hidden layer and a sigmoid activation function were obtained for total phenol content( =0.9539- =0.9865), antioxidant ( =0.9839- =0.9649) and vitamin C ( =0.9758); as for wide-edge loading in a network with 5 neurons in the hidden layer and hyperbolic tangent activation function,  the highest R2 rate of vitamin C content was obtained equal to =0.9865. According to the obtained results, the neural network with these two activation functions possesses an appropriate ability in overlapping and predicting the simulated data based on real data.