A blog post series about beer, using Collie!

[nathancooperjones]

The only thing better than beer is drinking beer, and the only way to drink beer is to make a decision and choose one. For me, this has always been a difficult decision, with so many categories to choose from before deciding on a single beer (dark or light, imported or domestic, can or bottle, a good beer or a Guinness, etc.).

To solve this existential crisis, I used Collie to recommend me beers based on my existing beer preferences. This three post blog series walks through the process of preparing data, training a model, evaluating a model, and training even better models, all using Collie.

Feel free to check out part 1 here (introduction, preparing the data, training a baseline model, and evaluating the model using MAP@10), part 2 here (training and evaluating better models using adaptive loss and multiple optimizers), and part 3 here (using metadata in the model and loss function for the ultimate model, and trying some of the recommended beers).

All code used throughout the blog post is available in the GitHub Gist here.

Cheers! 🍺