Clark: Humans are bad at performing 10-dimensional optimization mentally
Scott Clark · 705: With $8.8M Raised, Is This The Ultimate Machine Learning Tool? · Jun 29, 2017 · at 6:28
Scott Clark, co-founder and CEO of SigOpt, discusses why machine learning parameter tuning requires algorithmic optimization rather than manual intuition.
0:00 / 0:03exact quote · 3.6s
720p mp4 · rendered on demand · StarZero watermark
“Turns out humans are pretty bad at doing 10 dimensional optimization in their head.”
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 Scott Clark
Black-box Bayesian optimization outperforms standard machine learning tuning techniques
“What we're able to do is without any domain expertise, without making any assumptions about that underlying model, outperform these standard techniques by providing this ensemble of black box Bayesian optimization strategies.”
Algorithmic trading firms use SigOpt because client data never touches its systems
“This allows us to work with some of the most secretive algorithmic trading firms in the world, where their domain expertise and their models are literally how they make their billions of dollars, but they can still use SigOpt because all we're tuning are these…”
SigOpt fine-tunes existing client models instead of entirely replacing them
“So instead of just taking a raw data set of decades of fraud data and giving them some model to like rip and replace what they already have, we sit on top of what they have and provide this additive boost by fine tuning it.”