This BIG System (Small computers stay far away for your own good xD ) is able to recognise the feeling given be a smiley.
The current feelings the system can detect are "happy" and "unhappy".
The mechanism is based from the original "deep learning" system, modified by drawcoco to work using only integers.
It is made out of 2 layers, composed of 15 and 2 neurons respectivly. Each neuron do a sum of values (eather positive or negative) to end up with the "probability" the image "should be" of each feeling.
The system have been trained using the "metropolis" algorithm on computer using python.
Each neuron can handle values from -127 to +128 and works with signal strenght.
System done by drawcoco and RitzKid76, Asking both for Senior.
(I, drawcoco, am only regular but I think I deserve senior for all the core system done by me).
For more detail please contact drawcoco
Smiley must be created out of the 5x6 area.
You can see on the left of the screen the 9 smileys used to train the neural network. Any other smiley you create is "calculated" to fit the data the system have been trained with and try to guess the result (it can of course fail).
Actualisation time is about 30sec from start to finish.
This version is the V2 of a previous V1 that was about 1min delay.
For more info please contact drawcoco
Edit: Really sensitive to chunk unloading, this might interfere with the answer
Comments
little mistake
I am not learner but Regular, mistake in "Machine features"
little mistake
Fixed
- ub
On request denied
Could have been improved.
Improvement is being made.
New PR will be created.