a neural network capable of influencing it's own biases, simulating learning to a certain extent. currently pretty useless, but it an interesting proof of concept either way.
computer
neuron
circuit
electornics
electricity
neural
network
machine
learning
Comments
-
if your looking for other types of representaition like lights then you could traces lines from the outputs to a box that launches phot or uses bray or aray when triggerd and is made void that has a deco of black to make it look good. filt can be use to change the color of the phot.
-
for output monetering it depends what you want. if you want a gragph use a system that stacks when a certain output is made and then display it somehow (im'm not really good with theese so you may need to ask someone else about that).
-
srry if it's like a wall of text but there's lots of options
-
you may also remove the baccround surrounding the outer wiring. another option is to add frzz into some of the blank spaces of the neurons. you also may want to add to the number of backround pixels to increase the systems heat capacity, turning on ambiant heat will also cause a reduction in heating but it similarly reduces the effecy of cooling so it is up to you to pick the option that best fits the model you are using.
-
this will make the influences of the cooling increase while reducing the influence of the heating.
-
replace the backround filler with heat lines that lead to the cooling moduales, then delete pixels from the backround surrounding the heat moduales.
-
i have primitive biasing by privilege but its inconsistent also the output monitoring system is bad
-
2x the passive cooling so anything but active paths get purged?
-
zoidberg how do i implement a punishment system for a superactive network
-
yay fp!