How to run machine learning on a Mac SE 30 from 1989

This repo contains my work on getting a basic perceptron machine learning algorithm running on my vintage Mac SE 30. It was written in C using THINK C on System 7.1. The program contains the following features:

• A gui for the user to draw a letter or shape on a grid
• A perceptron algorithm that is trained to detect the capital letter "A"
• A console output that shows the perceptron's confidence level in detecting the letter "A" from the user's drawing

Requirements

• A Mac SE 30 running System 7.1, or an emulator like Mini vMac
• THINK C 5.0.2
• I recommend reading my last post on vintage Mac development here to understand how to setup THINK C

Layout of the project

The project consists of the following files:

• main.c: The main file that contains the parameters and orchestration of the gui and perceptron
• input.c and input.h: The logic and headers that contain the gui for the user to draw a letter or shape on a grid
• recognition.c and recognition.h: The logic and headers that contain the perceptron algorithm and training loop
• weights.dat: My pre-trained weights for the perceptron
• Machine-Learning: The THINK C project file

Running the project

With your project files in one folder, open Machine-Learning which should open in THINK C.

Make sure to add the full ANSI resource if it's not already there. We need the full one for this project to work. It will be in your THINK C install folder like Development/THINK C 5.0 Folder/C Libraries/ANSI.

Inside main.c you can change the highlighted code to make it train every time, or to look for the pre-trained weights.

Press CMD+R to run the project. You should see a window like this:

Fill out the squares into whatever shape you want by clicking on them.

Once you're done, hit Enter and the program will run the perceptron on your drawing and output the confidence level in the console.

Further Reading

I got the idea to try this out when reading about Optical Character Recognition (OCR) in these 50s-90s papers and articles:

• The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain - 1958
• Developer - The Apple Technical Journal - August 1992 - Page 85
• MacWorld - March 1985 - Page 67

They got me thinking about what was possible with technology back then, and I knew that it should be possible to do something simple like this on a Mac SE 30. I hope you enjoy this project and it inspires you to try taking a step back, slowing things down, and really looking at the foundations of simple technology.