Home : Projects : Lereco
I've been interested in machine learning for as long as I can remember. My interest in this subject condensed into something recognisably related to the current Lereco project around 1999, and most of the concepts behind the underlying model have been unchanged since around 2006.
Lereco 'machines' are software systems, constructed from fairly simple elements, which are capable of altering their own implementations on a fundamental level based on real-world experience. Based on a simple starting point, and running on moderately powerful off-the-shelf PC hardware, these machines can:
- Identify and react to stimuli
- Make decisions despite uncertainty
- Exhibit search-like problem solving behaviour
- Optimise behaviour based on experience
- Learn to interpret novel patterns
- Learn to plan and predict
I'm currently very happy with the model but of course as its creator I'm subject to confirmation bias! At the moment, as time allows, I'm working on software to help test and demonstrate the capabilities of this model. Once I feel I can show that these machines are scientifically interesting, I will start to consider the best way of going public with what I think is a really exciting discovery.