This meetup is a continuation of the last meetup on recommender systems held before the summer at Seal Software. We will now build upon what we learned last time, and extend the linear models discussed with non-linear models, by training deep recommender systems. The talk will be based on a Jupyter notebook and PyTorch. I will try to make it as self-contained as possible (without repeating too much from last time) so don’t worry if you weren’t at the last meetup!
Our hosts at Meltwater will also give a talk on:
A challenge of creating neural networks is to find the optimal network architecture. How many hidden layers should it have? How many neurons in each layer? Which activation functions should we use? It might be hard to reason about whats best for your problem and trying out all the different options is often too expensive or takes too much time. We will talk about how to approach this in a more efficient way using Genetic Algorithms for automatically learning a good network architecture.