Max/MSP Neural Network Tutorial 2: Abstracting the Neuron & Troubleshooting
6,484
Published 2017-11-03
Thanks to tons of feedback I got from Facebook and YouTube comments, we're able to make a couple of quality-of-life changes to the neuron patch. We then go on to abstract the neuron entirely, allowing it to be called up with a dynamic number of inputs.
As always, the patch/es can be downloaded from www.spearced.com
Head to the Facebook Max/MSP Group, join up and give them some love: www.facebook.com/groups/maxmspjitter/
Link to the article used from the previous video: medium.com/technology-invention-and-more/how-to-bu…
All Comments (9)
-
Thank you very much for having made all these tutorials, they are very enjoyable and clear. When you got to the end of the first tutorial and you had that "counter" moment, I smiled and thought "how many times have I had prolems with the counter object?" A very frustrating object to use beyond simple loops. Needs redesigning.
-
The way the patch is designed, will it remember “what it’s learned” every time it runs or would you have to train it every time?
-
Me during first video: I got this Me during this video: wat
-
How do you work around the fact that when you send a list of 0 0 0 the answer is 0.5 no matter what? Also, I tried using the training data below to have the neuron output 1 when the input is 1-4 (in binary) as this would be useful in creating a 1 dimensional cellular automata however with this data when I enter in 1 0 0 it gives me a 0 back despite there clearly being an instance of 1 0 0 outputting 1 in the training data. Is this because the neuron only figures out the one place in the list that's most important and kind of ignores the others? Is it capable of understanding what I'm trying to do? 0 0 0 0 0 0 1 1 0 1 0 1 0 1 1 1 1 0 0 1 1 0 1 0 1 1 0 0 1 1 1 0
-
How is this program called, and from where can I download it?
-
for the very right matrixctrl, isn't that the very left 5 columns are the inputs? and the very right column is output? I don't understand why "the very left 3 columns are correct and the rest are incorrect". can you explain? thanks!
-
thanks for these. why did you use [loadmess 1] instead of just using [gate 2 1]? the second argument would open the 1st outlet
-
These max patches are so hard to read (I use pure data myself) if you align the connections on top of each other. Why not just use sends and recieves. Also some other things are simpler in pd (not everything).
