transfer function purelin equation based on neural network toolbox

what is equation of pureline based on neural network toolbox? as i have read before, that purelin transfer function is a = purelin(n) = n. but from simulation in neural network toolbox get different result from the equation. is the equation wrong?

thank you.

[Brendan Hamm]

Your equation is correct, but in a neural network, n will be a linear combination of the inputs from the prior nodes multiplied by their weights and added to the bias.

So in effect purelin is just allowing this linear equation to pass to the next layer directly as opposed to applying another transfer function.

[Arrya Anandika]

yeah, i have done it. but when i compare matlab simulation with calculate equation, i don't get same result. what's wrong?

[Greg Heath]

How are we supposed to know what you did wrong if you don't demonstrate it???

Greg

Answer 1

[Greg Heath]

Read the documentation

help purelin

doc purelin

Greg

[Arrya Anandika]

i have read it. but in simulation with matlab and calculate with equation, its not get same result.

what is the problem sir?

[Greg Heath]

The weights are used on the transformed variables, not the originals.

Hope this helps.

Greg

[Arrya Anandika]

what it's mean sir?

can you explain it more? or with example?

thank you sir.

Anantastic

[Brendan Hamm]

The input layer will include a mapminmax transformation, which maps the minimum value of each input to -1 and the maximum value to +1.

[Arrya Anandika]

so, we must set input range -1 to 1 sir?

with normalization?

because, based example that i have tried, the target that i want is 0 or 1, but in equation, i got the result many variation, like 2, 3, even 23491.

i don't know what's wrong.

[Greg Heath]

1. Now and in the future,

SHOW EXACTLY WHAT YOU DID IN ADDITION IN YOUR ATTEMPT TO EXPLAIN IT!

2. Have you read the documentation and tried the examples?

Greg