PCPlus 288: The thinking behind neural networks

For the December 2009 issue I turned to the subject of perceptrons and neural networks. This one was very much written for the layman: I’ve never written a neural network in my life.

However the article did grow out of some ideas I’d been having about writing iPhone apps that make better use of the phone’s other hardware such as the camera, rather than just the accelerometer. I’d been very struck by this article about writing an app that, from a photo of a Sudoku board, could work out the grid and the existing digits and present the result on the iPhone screen. The author quietly tosses off that he implemented a small neural network to do the OCR (Optical Character Recognition) part. I should be able to do that!

And then just over the past day or so, I’ve heard about another cool OCR application called Word Lens, an application that not only recognizes text from a photo or the camera but can translate it too.

As I said in the article, AI tends to get a bad rap, mostly because of the over-inflated predictions its proponents tend to produce, but in reality things like OCR, speech recognition, face detection and recognition, and so on are changing our lives in subtle and cool ways.

This article first appeared in issue 288, December 2009. It was also published online at techradar.com.

You can download the PDF here.

  Loading links to posts on similar topics...