This dataset have 10 different attributes.
The CNN have three convolutional layers, with a MaxPooling of (2x2) and a Dropout of 0.25. Also, after the CNN it have two Linear Layers. The first one with 1024 input and 500 output, second one with 500 input and 10 output. The number of epoch for training was 30. The accuracy of this CNN was of 74%.
The CNN is the same as the first attempt but the number of epoch was increase to 50. The accuracy results was of 78%. If number of epoch are increase is possible to improve accuracy of the CNN.