fuzzy transformation; convolutional neural network; kernels; convolution; features;

We propose a new convolutional neural network ? the FTNetand explain its theoretical background referring to the theory of a higherdegree F-transform. The FTNet is parametrized by kernel sizes, on/offactivation of weights learning, the choice of strides or pooling, etc. It istrained on the database MNIST and tested on handwritten inputs. Theobtained results demonstrate that the FTNet has better recognition ac-curacy than the automatically trained LENET-5. We have also analyzedthe FTNet and LENET-5 rotation invariance.