Page 2 of 2

Re: Embedded AI

Posted: Sun Jul 26, 2026 5:14 am
by mnfisher
The type of 'w' looks to be incorrect.

It would need to be a 2 dimensional array of floats for the code to compile.

Code: Select all

Float w[10][10]
for example.

I would guess it is just an array and you are attempting to pass a single element to calculate.

Code: Select all

Float w[10]   - w[i] 
is a float not an array. (An oddity in the forum code -

Code: Select all

w[i]
displayed as just 'w' if not on a code block) - if this is the case the parameter would need to be w.

Please attach your code demonstrating errors like this (full editor attachments or drag and drop to the full editor window) as we can only guess what is happening without seeing the full code.

Martin

Martin

Re: Embedded AI

Posted: Sun Jul 26, 2026 9:44 am
by Steve-Matrix
If w[] is a one-dimensional array, just pass "w" as the parameter.

Re: Embedded AI

Posted: Tue Jul 28, 2026 4:58 am
by 17712885438
Why is there no output for this example z?
neuron1.fcfx
(16.05 KiB) Downloaded 179 times
(view online)

Re: Embedded AI

Posted: Thu Jul 30, 2026 11:14 am
by BenR
Hello,

How's this? I've tweaked your file slightly and it seems to be working now.
neuron1.fcfx
(17.13 KiB) Downloaded 200 times
(view online)

Re: Embedded AI

Posted: Sun Aug 02, 2026 6:38 am
by 17712885438
The problem has been solved. Thank you!

Re: Embedded AI

Posted: Thu Aug 06, 2026 1:52 am
by 17712885438
Could we add a macro command for calculating the loss? This will further enhance the functionality of the neural network components. Thank you.

Re: Embedded AI

Posted: Thu Aug 06, 2026 1:55 pm
by BenR
Hello,

Yes that's been added for you now. The latest version of the component is available from the Help -> Library updates menu now.

Re: Embedded AI

Posted: Tue Sep 08, 2026 12:13 am
by 17712885438
Neuron.fcpx
(4.21 KiB) Downloaded 11 times
.Could it add an option for the softmax activation function? thank you.

Re: Embedded AI

Posted: Tue Sep 08, 2026 9:15 am
by BenR
I'll have a look, thanks for the suggestion.

As per your other AI related request we're discussing options tomorrow.