machine learning - Is weights are diffrent for each training Example in perceptrons -
i new neural network. have training dataset of 1k examples. each example contains 5 features.
initially, provided value weights.
so, there 1k value stored weights associated each example or weight values remain same 1k examples?
i.e. example1 => [f1,f2,f3,f4,f5] -> [w1e1,w2e1,w3e1,w4e1,w5e1] example2 => [f1,f2,f3,f4,f5] -> [w1e2,w2e2,w3e2,w4e2,w5e2] here w1 means first weight , e1, e2 means diffrent examples or example1,example2,... -> [gw1,gw2,gw3,gw4,gw5] here g means global , w1 means weight feature 1 on. thanks
start single node in neural network. it's output sigmoid function applied linear combination of input shown below.
so 5 features have 5 weights + 1 bias each node of neural network. while training, batch of inputs fed, output @ end of neural network calculated, error calculated respect actual outputs , gradients backpropogated based on error. in simple words, weights adjusted based on error.
so each node have 6 weights, , depending on number of nodes (which depends on number of layers , size of layers) can calculate number of weights. weights updated once per batch (since doing batch training)
wiki

Comments
Post a Comment