Article Preview

FEATURE

Get Your Machine Learning On

Exploring ML with Xojo

Issue: 18.3 (May/June 2020)
Author: Jim Meyer
Author Bio: Jim is semi-retired. He enjoys traveling, sailing, and working on interesting projects. He started working with Machine Learning out of pure curiosity. (This article is an abridged version of the sessions Jim was to give at XOJO.CONNECT. At this time, he does plan to speak at the London 2021 event where he will go into more detail along with demonstrations using some of the hardware described in the article. He also expects to update the presentations with any new developments in the field.)
Article Description: No description available.
Article Length (in bytes): 23,073
Starting Page Number: 58
Article Number: 18305
Resource File(s):

project18305.zip Updated: 2020-05-01 11:31:16

Related Link(s): None

Excerpt of article text...

Machine Learning (ML) is no longer just found in science fiction books and movies. It is here today and are not only impacting the IT world, but many aspects of our daily lives. Understanding how ML systems are built and deployed is an important first step in helping you evaluate potential uses and misuses of this rapidly evolving technology.

What is ML?

As developers, we traditionally build systems that perform specific tasks. ML systems are different, they are designed to "learn" and then use that acquired knowledge to make predictions or decisions. Over the past several years, these systems have become so advanced that they are now considered a form of Artificial Intelligence (AI).

Learning is usually done using specialized algorithms that scan and evaluate large historical datasets. That process results in a model, which is then used to make the desired predictions or decisions.

Learning without a historical dataset is also possible, but in actuality such systems are just accumulating their own data. A good example is gaming where the learning is done by playing millions of games, accumulating more knowledge with each win or loss. The Alpha Go Zero project is one of the best examples (See https://deepmind.com/blog/article/alphago-zero-starting-scratch).

There is also a technique called "Transfer Learning" where an existing general purpose model is fine-tuned for a more specific use. Your phone uses this technique to identify people that you have named in your personal photos. The model provided with the phone already knows how to detect and identify individual faces and you are simply adding their name to the model.

...End of Excerpt. Please purchase the magazine to read the full article.