This article provides an adaptive framework to train and serve on-device recommendation system model. This approach personalizes recommendations by leveraging on-device data, and protects user privacy without having user data leave device.

The goal of this article (which happens to be my first Medium article 😊) is to share with you what I have learned while comparing the performance of three different deep learning models on the infamous NLP sentiment analysis task. We will be asking our models to predict positive (1)    

Below are my notes on Andrej Karpathy’s video tutorial on introduction to language modeling. You can watch Andrej’s original presentation on youtube. In Part 1, we worked on a bigram model that takes into account only the local context of a word. This approach is    

In this post, I will revisit in more detail our previous work that uses human-inspired likelihood revision or similarity to probability [Blok et al. 2003] to re-rank or score any word or text fragment based on the semantic relation to an external context. We will use the most popular Semantic Similarity pre-trained model (e.g., w2v, GloVe, fasttext, etc.) to compute these relations.