(Ln(x))3

The everyday blog of Richard Bartle.

RSS feeds: v0.91; v1.0 (RDF); v2.0; Atom.

Previous entry. Next entry.


1:05pm on Saturday, 1st November, 2014:

Training

Anecdote

This week's shopping came to much less than usual. The stars had aligned, and we hadn't run out of much at all. In fact, it was less than half what I'd normally pay. As I approached the scan-and-go payment point, I wondered if the Sainsbury's computer would spot that I'd bought less and as a result suspect that I might actually have bought the same but not scanned it all.

Sure enough, when I tried to pay an assistant was called over and I did a partial scan of my shopping. This is the first partial scan I've had; normally they scan the lot, but I guess if they think that I've neglected to scan half my shopping then it wouldn't take a whole rescan to ascertain that.

Hmm, so let's think about this. Let's say that I plan my shopping so that I stock up for three weeks then buy less in the fourth week. I'll pay slightly more for three weeks then a lot less for one week. The system computer will pick up on the fourth week and decide (perhaps at random) whether to order a rescan of my shopping that week. After a few of these, it will notice the pattern and adjust its model accordingly. It'll expect three large shops then a small shop, and only throw an alert if there's a change to the pattern that couldn't be explained by, say, Christmas.

Having thus trained the system, I could then buy a ton of stuff on the fourth week, only scanning the usual things I'd been scanning before, and get away without paying for the rest.

Then again, I wouldn't put it past Sainsbury's to have enough data on me to include the fact that I have a PhD in Artificial Intelligence and therefore may know about training neural networks, so it would hit me with a rescan anyway.




Latest entries.

Archived entries.

About this blog.

Copyright © 2014 Richard Bartle (richard@mud.co.uk).