The NIPS conference (Neural Information Processing Systems) is one of the most prestigious yearly events in the machine learning community. At each NIPS conference, a large number of research papers are published. Over 50,000 PDF files were automatically downloaded and processed to obtain a dataset on various machine learning techniques. These papers discuss a wide variety of topics in machine learning, from neural networks to optimization methods and many more.
This is a diagnostic analysis of Machine Learning topics from NIPS papers data using LDA, GridSearch from sklearn, and CoherenceModel from gensim.
In the first notebook - sklearn, I use the Natural Language Processing (NLP) techniques, Latent Dirichlet allocation (LDA) and GridSearch from python's sklearn library to analyze machine learning topics in the NIPS papers dataset. I use the concepts of perplexity and log likelihood to find the optimal number of topics for the data. I then plot the various models that come out of the GridSearch fit against their perplexity to see which models perform best.
In the second notebook - gensim, I again use LDA, this time with CorehenceModel from python's gensim library to analyze machine learning topics in the NIPS papers dataset. The concept of topic coherence measure is used to find the optimal number of topics for the data. Finally, I use python's LDA visualization tools to display the results of a model with the aforementioned number of topics.
Data taken from kaggle.com. Find the dataset here under papers.csv.