JosiahNet: live coding a GOLANG Deep Learning Library SourceCode: github.com/thenomemac/josiahnet Inspired by Joel Grus’s youtube live coding of a deep learning frameworks, I wondered as a very new to GOLANG user could I live code a deep learning framework? It ended up taking a few hours to code this up as I didn’t have numpy as a starting point, but I found GOLANG to be very suitable for implementing a Deep Learning Library in Go with no dependencies.
This tutorial walks users through an entire deep learning based NLP pipeline. The notebook demos a sentiment model for IMDB movie reviews using an LSTM and benchmarks the model against a linear model with tokenized TFIDF features. The LSTM model is then introspected by looking at works most likely to flip a review from positive to negative and vice-versa. Then TSNE is used to visualize the the distribution of the most impactful words in movie reviews.