date_range 20/11/2022 00:30

Here we summarize a paper from “Predictive Model Performance: Offline and Online Evaluations” from Microsoft. It analyzes the offline and online metric discrepancy problem, and the simulated metric for simulating online performance.

date_range 06/11/2022 00:30

Here we introduce a paper from Facebook about ads click through rate prediction. It was an old paper published in 2014, but I believe it still has lots of good practices which we could learn from for ads service.

date_range 29/10/2022 21:30

Here we introduce Gmail’s Smart Compose which provides real-time and interactive writing suggestions. It is a large-scale neural language mode of sequence prediction. We use the question-answer format here.

date_range 10/09/2022 22:30

Here we introduce a Youtube’s recommendation system based on deep neural networks. It consists of a deep candidate generation model and then a separate deep ranking model. Youtube recommendations have to deal with scalability, freshness and noises of user feedbacks. The two-stage of this model can get a small personalized and accurate recommendation for users from a large of millions of videos.

date_range 02/04/2022 17:30

In order to keep notes and share my understanding of computer algorithms, I summarize some of them here (the specific content will be continously updated). The example codes are in Python 3.

• Binary search

  • Binary search summary