Three most important lessons about neural networks and word embeddings: 1. No free lunch, 2. Size matters, 3. Engineering matters
With 2 friends of mine (Team was: Maksim Belousov, Mike Phuycharoen, Nikola Milosevic (myself)) on 12th and 13th November I participated at the GreatUniHack that was held in John Dalton building of Manchester Metropolitan University. We were implementing idea of machine learning-based music composer, for which we later came with name SchumaNN. For some time I wanted to experiment with this idea and this 24 hour hackathon came as a perfect match. We had couple of meetings prior to hackathon when we discussed some solutions and papers/blog posts we read. Some of the approaches we found are:
The decision we went at the end is to create a recurrent
…
