[REMINDER] Notes + some tiny typos

[GiorgosXou]

NeuralNetworks/examples/FeedForward_double_Xor/FeedForward_double_Xor.ino

Line 1 in dee00f6

#define NumberOf(arg) ((unsigned int) (sizeof (arg) / sizeof (arg [0]))) // calculates the number of layers (in this case 4)

*calculates the number of "elements in an array" not "layers"

[GiorgosXou]

In cases of any change to the examples, reminder to update:

• https://robotics.stackexchange.com/a/105773/37506
• https://www.reddit.com/r/arduino/comments/181gisb/resourceconscious_neural_network_implementation/

[GiorgosXou]

This softmax part-implementation for use with any ACTIVATION__PER_LAYER, (even thought I trust myself, not gonna lie) seems slightly suspicious:

NeuralNetworks/src/NeuralNetwork.h

Line 2041 in da5b5af

void NeuralNetwork::Layer::Softmax() {for (int i = 0; i < _numberOfOutputs; i++){outputs[i] /= me->sumOfSoftmax;}}

And by suspicious, I mean: i don't remember 😭 + division lol

[GiorgosXou]

Reminder: Remove space in middle after = :

NeuralNetworks/src/NeuralNetwork.h

Line 1509 in da5b5af

layers[i].weights[j][k] = *((DFLOAT*)(&tmp)); // toFloat() which is atof() is not accurate (at least on Arduino UNO)