question basic / machine learning

[debluzz]

would like to know if there is any machine learning software works with liberty basic. Or is there a way to make one with liberty basic.

[bplus]

There is a way, if there is a will!

[bluatigro]

i done some work on genetic programming somwere on this forum

[tsh73]

It really depends on what are you meaning under "machine learning"

There some old approaches that could be translated to LB
Like program that learns to tell apart different animals
justbasiccom.proboards.com/thread/779/program-tells-apart-animals

or Perceptron that learns to distinguish characters
en.wikipedia.org/wiki/Perceptron
(I've seen a few of such threads on old forum, could dig some -- not sure it actually works
actually very very long ago I did one on some 8bit basic, nothing left of cource
)

Any approach that could be programmed from scratch
and where performance of LB is not big issue (because you REALLY can run 100x faster with compiled language!)
could be tried.

But things like using big ML libraries python-style (just googled a few - SciPy, Scikit-learn, PyTorch)
- sorry, they just not exist for LB.

[Rod]

Let’s say this discussion is about learning to predict the outcome of a race, say a dog race since this is what we have discussed in the past.

We are really talking about something similar to a bank score card or credit rating. Something that predicts good or bad, win or lose.

These systems never predict accurately, they just segment the odds so you can choose a level of risk.

But the data used must be meaningful. Trying to predict a dog’s performance on the basis of past performance probably isn’t going to work.

You need more data, how fit is the dog today, how well fed, how well trained. So crucial info might be missing.

But I guess at the task.

However Liberty is well placed to build a learning system. But you have to instruct the computer how to learn, it cannot just learn itself.

If you want to progress you need to provide as full a description of the task that you can.

[Carl Gundel]

Feb 23, 2022 12:51:38 GMT -5 Rod said:

However Liberty is well placed to build a learning system. But you have to instruct the computer how to learn, it cannot just learn itself.

I guess that some sort of simple but useful machine learning example would be good for Liberty BASIC because schools seem to be interested in machine learning.

[Rod]

Well in the past we have learned how to play tic tak toe. Recently we have had a hoover learn how to clean a room. The computer learning how to tell animals apart. The computer playing chess!

Perhaps we need a new board to focus attention on machine learning examples.

[Carl Gundel]

Feb 24, 2022 3:28:27 GMT -5 Rod said:

Well in the past we have learned how to play tic tak toe. Recently we have had a hoover learn how to clean a room. The computer learning how to tell animals apart. The computer playing chess!

Perhaps we need a new board to focus attention on machine learning examples.

It seems like we need a definition of machine learning.

[irvbingham]

Definition:
Machine learning is the study of computer algorithms that can improve automatically through experience and by the use of data. It is seen as a part of artificial intelligence. Machine learning algorithms build a model based on sample data, known as training data, in order to make predictions or decisions without being explicitly programmed to do so. [Wikipedia]

[Brandon Parker]

Here's the ISO definition: ISO/IEC 2382-31:1997(en)

Arthur Samuel said that ML is a “field of study that gives computers the ability to learn without being explicitly programmed.”

So, in generalized terms, rather than the programmer producing a specific algorithm to analyze data, the algorithm itself is determined by the data presented, and the system can then "learn" from that data and reshape how the entire system works. That is at least how I tend to generalize it in my head.


{:0)

Brandon Parker

[bplus]

ML is branch of AI that "learns" patterns from provided examples. Perceptatron is good example, guess Animals and chess not so much, those are different kinds of AI. TTT might be depends on how it's programmed, do you provide it with examples of winning and loosing games (ML) or do you say in code, if such and such then do this...

Dang 2 posts show up while I am typing out my rely after Carls, great minds think at the same time LOL.

[Rod]

Fishers 1936 dataset on iris identification might set a challenge for us. It is a small dataset often used for ML projects. Oodles of info available to help understand the dataset.

[Rod]

Actually Fishers data set has its opponents for various reasons. This might be a more fun alternative.

allisonhorst.github.io/palmerpenguins/articles/intro.html

[tenochtitlanuk]

Blatigro and I have played with some AI stuff. See my website for things like fuzzy logic; k means; etc.

The Fisher/Iris database was one of the things I played with. It is regarded as a salutary example to beginning AI programmers because the data contains three species of iris, but you can only distinguish one from the grouped set of the other two. If I find time I'll try it on the penguins..

I was particularly interested because my father was a Cambridge mathematician in the Thirties. He was R W Fisher, and every so often people with a maths background have asked 'Was he THAT R Fisher?'. Sadly not, but he was a much better mathematician than me! I'm much more interested in science/technology, and struggled with the abstractions of high-level maths. ( that's UK-speak for what my US friends call 'math'.)

[Rod]

I have looked at a few tutorials now. It's kinda funny. Despite the definition stating that the computer should learn, most examples start off with a slew of maths and statistical analysis that a human needs to learn. They then move on to strategies, err..... isn't that the computers job?