Neural Nets - Prof. Mark Whitehorn

[This is a first-come, first-admitted event. In the very unlikely event that more people turn up than the building capacity of 100, then we will not be able to accommodate those who turn up after we have admitted 100 people.]

Neural Nets are gloriously paradoxical. On the one hand, I can very easily (and will, during the talk) explain to you how they work. On the other hand, suppose I build a neural net to solve a specific problem (say, loan approval) and you ask me “Why has your neural net approved this loan but not this one?” I have to reply “I dunno.”.

It is a form of double-think; we simultaneously do, and do not, understand neural nets.

The good news is that, by the end of the talk, I hope you will be capable of the same trick; understanding both how they work and why it is so hard to explain why a given net makes a specific decision.

There will also be time to talk about how stonkingly useful they are and how powerful they have become.