|
This has been replaced with the C# version here: https://github.com/NTDLS/AIVolution DetermiNet is an experimental development in the field of artificial intelligence - an eventual component of a soon to be released (incomplete, to the open source community) multi-generational, evolution and machine learning project known as: AIEvolution.
DetermiNet is a C++ implementation of a multilayer perception neural network with two outside-the-box goals in mind: (1) seamless ability to merge neural network's to combine decision making capabilities and (2) prove/disprove the academic question of multi-hidden-layer benefit. Summary Operation
^ I did this in MS paint-brush. Yea, that's masochism baby! ^
Prediction Methodology#1 Supplying data (quasi-boolean)
A value is supplied to each of the 3 input nodes (F) and the passed in values are matched to known input states (I).
For example, if the value of "Has Children" is "Yes" then a value of 1.0 is passed from the input state node (A) to each of the input interface nodes of the hidden layer (D).
If the value of "Has Children" was passed in as "No" then the value of 1.0 is passed to the input interface nodes of the hidden layer (D) by the input state node (B) and a value of 0.0 is passed from the input state nodes (A) and (C) to the input interface nodes of the hidden layer (D).
The approach in this example is repeated for each of the three input nodes (F). If a value is not supplied or if a supplied value does not match any of the known input states (I), then a value of 1.0 will be supplied to the input interface nodes of the hidden layer (D) by the "missing state" node (H) and 0.0 will be supplied to the input interface nodes of the hidden layer (D) by all of the other nodes in the given input state cluster (I).
#2 PredictionSome magic happens. To be continued…
#3 ResultsThe output node at the output node layer (J) with the highest value is the node whose value is predicted.
|