Pyhton, Keras, SciKit-Learn, Matplotlib: Machine learning research project around classification of intent behind tech support emails in order to enable automatic follow up.
Optimisation of hyperparameters (epoch, batch size etc):
This seems like a relatively straightforward optimization problem, and there is lots of advice available.
Optimisation of network topology
There is not a lot of advice in this area, and I am unsure how to move forward with this. I could create many iterations with different numbers of layers, dropout, pruned connections, width of layers and so forth, but this feels very inefficient. Do you know of any good methods?
Hyperparameters - batch size / epochs https://towardsdatascience.com/simple-guide-to-hyperparameter-tuning-in-neural-networks-3fe03dad8594
Learning https://towardsdatascience.com/neural-network-optimization-7ca72d4db3e0
Topology / Architecture
Drop out Embeddings normalization