Can NEAT produce neural networks where inputs are directly (without intermediate hidden neurons) connected to outputs?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
Yes, it is possible (depending on the nature of your problem), using the four types of standard NEAT mutation, but it is improbable.
When the NEAT algorithm begins, it operates on a blank canvas. After each generation, the algorithm will either:
Construct a new axon
Construct a new node on an existing axon
Update existing weights/bias
Remove a node or axon from the network
(However, in general, NEAT does not produce neural networks where two input (or output) nodes are connected.)
-
$\begingroup$ Thank you I wanted to ask what you put in the edit section of your answer. But can we say, "outputs and inputs are nodes like other Neurons"? They don't have activation functions I believe, but two or more middle neurons can be connected to one output, no? $\endgroup$ Jan 19, 2019 at 17:29
-
$\begingroup$ input nodes are just numbers that are inputted into the system, and do not have activation functions because they are only processing one number. They do, however, feed the input they received to all of the nodes on the next layer through a series of weighted axons. Those weights are processed by the layer receiving the inputs. The output neurons have activation functions because they combine the output of all the nodes in the middle layer into one output for the system. All the middle neurons are always connected to all the outputs (generally). $\endgroup$– iamPresJan 19, 2019 at 23:28
-
$\begingroup$ Thanks, what you mean by; All the middle neurons are always connected to all the outputs (generally). I know it happens when you are working with Deep Neural Networks. Does this happen in Neat ? Isnt it random process? $\endgroup$ Jan 20, 2019 at 10:09
-
$\begingroup$ Yes. In neat the process is random and nothing needs to happen. $\endgroup$– iamPresJan 20, 2019 at 17:49