User:Sedenstrand/Malmberg neural net circuit
This is one of the fastest way to simulate a neuron. One neuron every clock cycle. It´s based on the model of an artificial neuron and is about 600 times faster then a computer at simulating neurons to the same cost and needs no CPU. The speed mutch depends on the number of RAM memories.
Function[edit]
The weights represents by voltage. The binary representation of a weight enables resistance to ground that together lowers the voltage. Output is 1 if Volt(weights) > Volt(ref).
See also[edit]
- Artificial neural network
- Perceptron
- Memristor ,developed by Greg Snider of HP Labs
References[edit]
- Information and communication institute(2010), [1]