Week three and the focus is firmly on machine learning, how we teach computers to think and distinguish between categories, messages, and obstacles.
Essentially, its how Skynet works. At least when the robopocalypse comes, I’ll know how they reached the decision to classify me as ham…
Joking aside, this has been the most interesting element of the course so far. Not only in terms of the taught materials, but it feels like we’re getting to know the teaching staff a little more.
Prof. Thrun was one of the team behind the DARPA Grand Challenge winning vehicle Stanley. Stanley is an autonomous vehicle, which had to navigate a preset course in a specified time.
See below to see it in action.
Prof Thrun was also involved in Google Street View, both stories were used to highlight how machines distinguish between different categories of object. In the case of Stanley this might be the difference between the road and the verge.
The section on distinguishing between different body types was also fascinating.
The units concentrated on supervised and unsupervised learning.
The course then looked at the mathematical principles behind this work. It ranged from common sense, to brain aching levels of algebra (at least for me).
This week we’re due to study Markov models and Bayes Filters. I can honestly say, I’m still enjoying it, still wishing I’d paid more attention to maths in school, but still loving every minute.