Random Forest AI from Weka on an Uno

[StAugustine]

I've used Weka (a machine learning suite of programs) to generate a Random Forest algorithm to accomplish speech selection on 10 different sounds. I would like to put onto an Uno. I have no idea of how to program the Uno with a Random Forest algorithm. I've searched the Web. Apparently transferring from Weka to any other machine is difficult. Hope you have some suggestions. Bob

[jremington]

How would speech samples be stored in the tiny memory available on the Uno?

[StAugustine]

Hi:

As I see it the random forest is a set of a couple of trees. The trees would be the decision makers and there is no need to store the speech samples. The Weka Random Forest algorithm has generated a model and I don’t know how to read it. I would like to go from the weka random forest model to C. If I used the J48 tree algorithm in Weka it is easy to write a C program for that. I’m working off on Instructables article on identifying speech and with its author to get my speech for the machine learning part. .

Bob

Robert (Bob) C. Gillespie

904-540-7679

"There is no such thing as spare time, no such thing as free time, no such

thing as down time. All we’ve got is this life time. Go. [Be]" — (H.L. Garfield)

[Klaus_K]

@StAugustine

I would like to put onto an Uno.

I am afraid the Arduino Uno is really bad a math. It might be able to add a few numbers faster than you and me but in the world of machine learning it would not be allowed to the starting line for a competition.

Have a look at the following blog entry about Arduinos first step into the world of machine learning.

https://blog.arduino.cc/2019/10/15/get-started-with-machine-learning-on-arduino/

The Arduino Nano 33 BLE uses a ARM Cortex-M4 processor which seems to be the lowest entry point for this kind of applications. You will need a processor that is at least as powerful as that.

[StAugustine]

Exceptionally interesting. I will study it. Here is the link to the article in Instructables. It is also an experiment and a learning experience. I’ve learned a great deal using the data from the Instructables article and that is why I’m pursuing the potential for installing a Random Forest implementation on an Uno. Thanks for this reference. It is fantastic.

Bob

Robert (Bob) C. Gillespie

904-540-7679

"There is no such thing as spare time, no such thing as free time, no such

thing as down time. All we’ve got is this life time. Go. [Be]" — (H.L. Garfield)

[StAugustine]

Hi:

As I see it the random forest is a set of a couple of trees. The trees would be the decision makers and there is no need to store the speech samples. The Weka Random Forest algorithm has generated a model and I don’t know how to read it. I would like to go from the weka random forest model to C. If I used the J48 tree algorithm in Weka it is easy to write a C program for that. I’m working off on Instructables article on identig

Robert (Bob) C. Gillespie

904-540-7679

"There is no such thing as spare time, no such thing as free time, no such

thing as down time. All we’ve got is this life time. Go. [Be]" — (H.L. Garfield)

[StAugustine]

Thanks for this info. I look forward to studying it and learning. That is all I’ve done since I started this effort after reading the Instructables article. I am learning and I’m enjoying it.

Bob

Robert (Bob) C. Gillespie

904-540-7679

"There is no such thing as spare time, no such thing as free time, no such

thing as down time. All we’ve got is this life time. Go. [Be]" — (H.L. Garfield)

[jremington]

The Uno can easily run a simple neural network, with a hidden layer and 20 nodes or so, including the training and back-propagation step.

This example shows how to teach the Uno to recognize decimal digits, as they might be displayed on a 7 segment LED.

The main limitation is the 2048 bytes of RAM memory.

Note: A more complete explanation of the Arduino neural network code linked above is presented by the actual author of the code, in this blog page.

[StAugustine]

This is great!

Bob Gillespie
904-540-7679 Mobile
"Fewer invasions and more equations"