Abao

Can anyone explain how does mlp object works? if I wanna make one audio buffer learns from the other audio buffer in real time, what should I put in the right inlet and left inlet? and don't really get what does the "tlu" and "layer" function in the coll object? thanks!

Can anyone explain how does mlp object works? if I wanna make one audio buffer learns from the other audio buffer in real time, what should 

mlp-in-ml·star

Are you familiar with how a multilayer perceptron works? That's what ml.mlp implements. It's not a machine learning algorithm that can be trained in "real time."

Are you familiar with how a multilayer perceptron works? That's what ml.mlp implements. It's not  a machine learning algorithm that can be trained in "real time."

No, I am not, I am reading the wiki page again and again... just to know if theres a easier way to understand how it works rather than the theological way.

Not really. by my understanding, the mlp object take learning values from the left inlet and can be played by the right inlet, but I don't understand how the "layer" function take effect in this process and what's going on in the coll object. I have read about the wiki page a bit and have few understanding about why the layer function helps for recognition purpose.

The documentation of ml.mlp is a bit thin. A mlp is essentially a function simulator. It creates a "model" that approximates a function that would give you a specific output based on a specific input.

A multilayer perceptron needs to "train" before it gives the results you want. It trains on a set of data where inputs are associated with desired outputs. So, if the object gets values "x" and "y", you might want the output to be "a" and "b". That's what goes into the right input. The sub-patch is pretty confusing. It is sending in 5 value lists (two input values and three output values). It keeps doing that (training) until the error is less than a threshold (0.01).

This is where you can interact with a trained model. That is, after the ml.mlp has gone through the training above. This inlet cause the object to predict or generate output values based on the input values that you send it. It's not very useful until the training is done.

Once you've trained a mlp model, this coll is a way to save the result. I don't think that the actual data in the coll are supposed to be particularly human readable. You can read up on how mlp's work. Theses data represent the connections, weights, and offsets for each node and layer. Here's a nice explanation:

The right inlet:
A multilayer perceptron needs to "train" before it gives the results you want. It trains on a set of data where inputs are associated with desired outputs. So, if the object gets values "x" and "y", you might want the output to be "a" and "b".  That's what goes into the right input.  The sub-patch is pretty confusing. It is sending in 5 value lists (two input values and three output values). It keeps doing that (training) until the error is less than a threshold (0.01).

The left inlet:
This is where you can interact with a trained model. That is, after the ml.mlp has gone through the training above. This inlet cause the object to predict or generate output values based on the input values that you send it. It's not very useful until the training is done.

The coll:
Once you've trained a mlp model, this coll is a way to save the result.  I don't think that the actual data in the coll are supposed to be particularly human readable. You can read up on how mlp's work. Theses data represent the connections, weights, and offsets for each node and layer. Here's a nice explanation:
http://playground.tensorflow.org/ 

I wish the help file explained the arguments. My guess is that [ml.mlp 3 2 1 8] creates a mlp with 3 inputs, 2 outputs, and 1 "hidden" layer with 8 nodes.

It also doesn't say what activation function is being used. Any individual node can only reproduce a scaled version of that function. So, adding more nodes creates a network that can simulate a more complex function. The cost is longer training time. Adding more layers also can create a more expressive mlp. In my experience, though, the training cost for more layers is very high and the results aren't really much better.

It also doesn't say what activation function is being used. Any individual node can only reproduce a scaled version of that function.  So, adding more nodes creates a network that can simulate a more complex function.  The cost is longer training time.  Adding more layers also can create a more expressive mlp.  In my experience, though, the training cost for more layers is very high and the results aren't really much better.

In spite of what the ml.star overview says: this is supervised machine learning.

Thanks for the explanation! they have just updated the ml library(still with thin description), and I also found this one from Nick Gillian, do you know is this any different then the inbuilt ml.star one? in this one you can send message "help" to mlp object to read some details of the commands intro.

Thanks for the explanation! they have just updated the ml library(still with thin description), and I also found this one from Nick Gillian, do you know is this any different then the inbuilt ml.star one? in this one you can send message "help" to mlp object to read some details of the commands intro.

https://github.com/cmuartfab/ml-lib

I owe almost all my creative successes over the past few years to the generosity of the MAX community on these boards. Thanks to all. I am NOT a good max user, but I'm a dedicated max student, and I promise I'm always doing the best I can to keep learning.

This is my first time posting and It's terribly long and wrong I'm sure. Also, This might be the wrong place to post this question, perhaps I should start a new thread? If there is a better posting strategy for this Q pls let me know. Final note/apology: the Q may contain incorrect assumptions due to my woeful lack of experience with the poly~ object 

(and this may not be an ML question at all, but a poly~ question...

 I'm doing my simultaneous due diligence to understand poly~.

"How do i get midi or freq. data out of ml.som.spectral that I use to control other synths (pitch/harmony)?"

I'm trying to have the ML.som/spectral object control another synth

 (not the poly~ contained within the program). Primarily because I want more control over timbre than the built-in poly~ offers. Fundamentally, that's my goal: to get useful data out of ml.som.spectral to control other synths with sends. In this case, I'm defining 'useful' as either midi pitch data, or freq. data.

Popping open the poly~ I see that when I change x/y values by moving through the SOM, it affects the cycles's gain only. There is an mtof which runs to the cycle~ object's frequency, but monitoring its output, the freq. doesn't seem to shift as I move around the SOM. When I saw the mtof I assumed there was both live "M" and "F" activity I could monitor and send changing values from. But I see no midi info preceding the mtof, and the freq. it returns doesn't change. My only conclusion (and this is where my total ignorance of poly~ may be misleading me) is that the SOM is managing a set of 48 'fixed' freq. cycle~ objects, and 'chords' are produced by varying their relative gain...? If so, is there any way of accomplishing my general mission of 

locating changing midi data or frequency data in the poly~ patch that I can harness/syphon/send to control pitch harmonic activity of a seperate synth.

That said, I'm still at a loss about where in the ML.som.Spectral patch to find midi or freq. data that I can 'send' elsewhere.

PS - my plan is to try and post this on other threads that seem relevant (this may violate cycling's policy anyway, but even if it doesn't and is just poor practice/etiquette I apologize in advance).

I owe almost all my creative successes over the past few years to the generosity of the MAX community on these boards.  Thanks to all. I am NOT a good max user, but I'm a dedicated max student, and I promise I'm always doing the best I can to keep learning.

This is my first time posting and It's terribly long and wrong I'm sure. Also, This might be the wrong place to post this question, perhaps I should start a new thread? If there is a better posting strategy for this Q pls let me know.  Final note/apology: the  Q may contain incorrect assumptions due to my woeful lack of experience with the poly~ object (and this may not be an ML question at all, but a poly~ question...) I promise  I'm doing my simultaneous due diligence to understand poly~.

"How do i get midi or freq. data out of ml.som.spectral that I use to control other synths (pitch/harmony)?"  I'm trying to have the ML.som/spectral object control another synth (not the poly~ contained within the program). Primarily because I want more control over timbre than the built-in poly~ offers. Fundamentally, that's my goal: to get useful data out of ml.som.spectral to control other synths with sends.  In this case,  I'm defining 'useful' as either midi pitch data, or freq. data.

Popping open the poly~ I see that when I change x/y values by moving through the SOM, it affects the cycles's gain  only. There is an mtof which runs to the cycle~ object's frequency, but monitoring its output, the freq.  doesn't seem to shift as I move around the SOM.  When I saw the mtof I assumed there was both live "M" and "F" activity I could monitor and send changing values from.  But I see no midi info preceding the mtof, and the freq.  it returns doesn't change. My only conclusion (and this is where my total ignorance of poly~ may be misleading me) is that the SOM is managing a set of 48 'fixed' freq. cycle~ objects, and 'chords' are produced by varying their relative gain...?  If so, is there any way of accomplishing my general mission of locating changing midi data or frequency data in the poly~ patch that I can harness/syphon/send to control pitch harmonic activity of a seperate synth.

That said, I'm still at a loss about where in the ML.som.Spectral patch to find midi or freq. data that I can 'send' elsewhere.  

mlp in ml.star