Скачать с ютуб Multi-label Classification with scikit-learn в хорошем качестве

Multi-label Classification with scikit-learn

The challenge: a Kaggle competition to correctly label two million StackOverflow posts with the labels a human would assign. The tools: scikit-learn, 16GB of RAM, and a massive amount of data. The goal: place above 50% in a Kaggle competition against data scientists from around the world from the comfort of my laptop. The talk: lessons learned from going deep with scikit-learn for tackling a very tricky machine learning problem and dealing with a lot of strange text and many labels. Explore the wonders of tf-idf, multi-label SGD classification, the power of n-grams and developing intuition around feature design, along with spinoff applicability to other work Cerner is doing. About the Speaker: Chris Finn is a Senior Principal Architect and Distinguished Engineer in Cerner's Medical Informatics group. Since joining Cerner in 1991, he has worked on a number of R&D efforts at Cerner including semantic search, community e-prescribing, and most recently, research into machine learning topics involving textual analysis aimed at improving documentation quality. In addition to R&D responsibilities, Chris contributes to a variety of talent development and outreach programs, including contributing curriculum to the new Project Lead the Way computer science course being piloted across the country during the 2013-14 school year, as well as building out a DevArc Academy course on the topic of modeling and simulation. This talk was given at DevCon, Cerner's internal engineering conference. Check us out at http://engineering.cerner.com/ Cerner DevCon 2014 June 3, 2014