Josh Bloom: Data-driven approaches do not require prior theory to predict outcomes
Josh Bloom · Supernovas and Novel Insight: Where Machine Learning is Headed Next · Jul 15, 2017 · at 10:03
Josh Bloom, UC Berkeley astronomy professor and co-founder/CTO of Wise.io, explains how high-dimensional machine learning models bypass traditional theoretical modeling by learning directly from historical patterns.
0:00 / 0:21exact quote · 21.8s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“Well, the point is, and this is sort of the conceit of all data-driven approaches, is that you don't need to have a theory about why something's gonna happen. The idea is that you've got enough data, both in, in terms of the number of examples, and then also sort of the dimensionality of the data, that you can train a very, ah, good grad student, or you can train a very good algorithm to learn from the past”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Josh Bloom
Machine Learning Tools Have Neglected Time Series Data
“Now you're getting into some interesting, ah, places where machine learning hasn't spent a lot of time, which is on time series data. And what we wound up realizing in our own, sort of, domain specific research is that there weren't a lot of tools for us from …”
Bloom: Machine learning must focus on real-time future data over history
“In some sense that's the greatest imperative and like the gauntlet that I lay down in front of anyone is that you're not doing machine learning because it's cool and it's fun and you can learn something about the data from the past. You're trying to really use…”
Bloom: ML model detected supernova in 11 hours, driving Nature publications
“So one of the great things is our, ah, machine learning algorithm and framework wound up finding a new supernova that was in a very nearby galaxy. And because it was found about 11 hours after explosion, which were days earlier than had ever been found for tha…”
Bloom: Machine learning software acts like virtualized graduate students at scale
“Because the software exists that can actually sift through and look at that data as if it's, you know, essentially virtualized graduate students with a huge amount of domain knowledge, and do this at scale it allows you to take more and more data.”
Josh Bloom: Real-time astronomy ML know-how directly transfers to enterprise problems
“The actual insight that we wind up sort of learning how to do with astronomy data in real time on noisy streaming data is exactly that sort of same know-how that we wind up applying to more of these conventional problems.”