#Mercari

Introduction This is my third post in a series of six, exploring deep learning with structured and unstructured data with the FastAI library. These are the links to my earlier posts on data preparation and structured data model. In this post, I’ll be talking about language models (LM) and how I built a custom language model using the data from the name and item_description columns in the Mercari dataset using a pre-trained language model provided by FastAI. [Read More]

Introduction This is my second post in a series of six exploring deep learning with structured and unstructured data with the FastAI library. Be sure to check out my post on data preparation. In this post, I’m going to describe my efforts in building a deep learning model that only uses structured data. Much of the material here, including code and ideas, are taken on FastAI’s notebook on tabular data with the Rossmann Store Sales Kaggle dataset and the paper titled Entity Embeddings of Categorical Variables by Cheng Guo and Felix Berkhahn. [Read More]

Introduction Data comes in various forms such as images, text, and tabular form. Deep learning can be applied to each of these areas and has excelled by giving state-of-art results. In this blog post series, I’m going to explore how to apply Deep Learning to a mixture of data groups, specifically, text data and tabular data. This is part of a bigger research project that I’m working on, which uses medical data (excluding images) which often consists of different types of data. [Read More]