CodeReclaimers / neat-python

Python implementation of the NEAT neuroevolution algorithm
BSD 3-Clause "New" or "Revised" License
1.42k stars 493 forks source link

Bias value in the add node mutation #168

Open FaydSpeare opened 5 years ago

FaydSpeare commented 5 years ago

When adding a new node between two existing nodes the connection into the new node has a weight of 1 and the connection out of the new node assumes the weight value of the connection that was just split. As far as I'm aware, this is to increase the complexity of the network without changing the output just yet. However, the new node will be affected by the bias. So when a new node is added, shouldn't the bias of that new node be set to 0?