ScratchData LogoScratchData
Back to _paperN1's profile

ANN With Bias Nodes

_P_paperN1•Created May 29, 2017
ANN With Bias Nodes
18
12
179 views
View on Scratch

Instructions

Neural network approximating the trend of 50 (randomly generated) blue dots. Press space to create a new set of dots. Press space twice to create a different set of dots. Press t to see the neural network in action. Press s to stop everything.

Description

Demonstration of a neural network with bias nodes (nodes with a fixed nonzero value) to allow a non-zero output of the network from a zero input. Visually, this allows the function of the neural network to no longer pass through the origin. This was a sequel to my other neural network approximating a function, expect it uses the bias nodes instead of adding two different numbers for each synapse to perform a linear function on the neuron it receives (the numbers m and b in the equation y=m*x + b, where x is the neuron and y is the value the synapse gets and transfers to the next neuron). However, this project is still very slow, due to the 3D rendering and higher amount of synapses needed for more input neurons.

Project Details

Project ID163670156
CreatedMay 29, 2017
Last ModifiedMay 14, 2018
SharedJuly 6, 2017
Visibilityvisible
CommentsAllowed