— Insight —
It's natural for data scientists to be interested in technical skill development but oftentimes, successful data science projects require a different kind of understanding. This post in the Harvard Business Review explores common issues and offers practical suggestions for developing your organization's data science proficiency.
This post by Ann Spencer recaps a recent talk by Pete Skomoroch titled, "Executive Briefing: Why managing machines is harder than you think." The focus is on the product management part of the talk, which includes discussion about how ML products are different than typical software products, traits of a good ML product manager, ML product development, data quality, and testing. There are a lot of useful insights here.
— Profiles —
Mike Schroepfer is in a position he never wanted to be in. As Facebook's CTO, he's responsible for curing Facebook's toxic content issues and that job comes with a bit of stress. This is a great read about his career, AI practicalities, culture at Facebook, etc.
— How-to —
Recent research suggests that semi-supervised learning is starting to work. In this post, Vincent Vanhoucke, Principal Scientist at Google, describes how the latest advances could be a huge development for machine learning. Includes links to key papers.
— Resources —
Great curated list of decision, classification and regression tree research papers with implementations from major conferences.
— Data Viz —
Nice interactive from the team at Figures. This is an educational piece that shows how airports are essentially wind maps, made from steel and stone.
Interactive articles like Trails of Wind in the link above are popular but is that enough? They take a lot of work to create and their effectiveness has not been widely studied. In this paper, researchers from the University of Washington's Interactive Data Lab introduce a toolkit that helps determine how well an interactive visualization is able to deliver information.