Machine Learning in the MAIL.RU Advertising System
summary
MAIL.RU’s advertising system uses machine learning on clickstream data stored in HDFS (about 100K records/sec) to train classifiers that segment Internet users. Features are represented with a high-dimensional sparse bag-of-words model, and dimensionality reduction via Latent Dirichlet Allocation (LDA) is applied, which in some cases also provides topic modeling of the feature distribution.
AI-written summary · about
related talks
33mMachine Learning to moderate ads
41mChallenges of Machine Learning at Scale
56mMachine Learning in E‑commerce
27mLarge Scale Response Prediction to Serve Ads on LinkedIn
19mMachine Learning for Privacy-Preserving Ads
49mThe Impact of Machine Learning in Companies
37mAutomated Machine Learning: Building a Pipeline
53mReal-Time Big Data Analytics with R