This is my magister's project on that with I was recearched how to choice the best architecture of dense neural network

The abscissa and ordinate will be the number of layers (1 - 8) and the number of neurons in the layer (1 - 64), respectively, and the applicate will be equal to the obtained quality score of this architecture for the selected activation function. In addition to the surface of the average test results of each activation function, also obtained its smoothed appearance by B-spline interpolation from the scipy package, giving the abscissa axis more partition to form a 64 × 64 rectangle. Also display a smoothed and original surface from above using a pseudo-color graph from the matplotlib package