All Posts

Ad is the major monetizing product for many companies. Here I'll introduce how I build the simulator to evaluate and develop the ad system in indeed.

Item Embedding is the new standard for recommend-system. Here I'm gonna introduce how I use both Item2vec and Siamese Network to build recommendation models learn from both item content and user behavior.

TextCNN is my new standard for quick-and-dirty fast NLP model building, it's fast and simple, natively include feature engineering as model and trained together.

Here I will try to explain some algorithm and implementation details about the work "the tensroflow chatbot" in layman's terms. Like dictionary space compression/projection, anti-language model, reinforcement learning... etc.

A deep-learning chatbot with (seq2seq model + attention mechanism + beam_search algorithm + anti-language model) in tensorflow, works end-to-end from training corpus to chat model, and build-in a facebook-messenger backend server.

Scraping twitter content from twitter streaming API, in python3.

A minimum facebook messenger backend in python flask, with a much simpler guide through how to create your facebook messenger app and setup!

Docker containers for machine learning and deep learning GPU support.

My code for Facebook Challenge: "predicting user checkings".
The goal of this competition is to predict which place a person would like to check into. For the purposes of this competition, Facebook released 40M check-in data of a small city, and our task is to predict the most likely check-in places of 8M user samples.

It's my code doing the Data Kaggle challenge. In this challenge, we are requested to develop a system predicting whether an article is a native advertisement. It’s involves lots of natural language processing tricks, like boilerplate removing, topic modeling and embedding techniques.

My content management system written in Ruby on Rails