rrherr

Reminds me of this:

"Data Scientist (n.): Person who is better at statistics than any software engineer and better at software engineering than any statistician."

https://twitter.com/josh_wills/status/198093512149958656

I've also found these to be helpful:

The Language and Grammar of Mathematics, from The Princeton Companion to Mathematics, by Timothy Gowers: http://press.princeton.edu/chapters/gowers/gowers_I_2.pdf

Reading Mathematics, by John Hamal Hubbard: http://www.math.cornell.edu/~hubbard/readingmath.pdf

"I buried the lede. It’s an in-browser reactive JavaScript notebook for exploratory analysis, visualization, and explanation." https://twitter.com/mbostock/status/858018259907620865

Fascinating topic. I wish I knew how to interpret Coltrane's diagram. Any ideas?

I recommend reading "Do machines actually beat doctors?" by Dr. Luke Oakden-Rayner, who is both a radiology doctor and deep learning researcher.

https://lukeoakdenrayner.wordpress.com/2016/11/27/do-compute...

Fascinating and educational to see all those variations! Reminds me a little of 99 Ways to Tell a Story by Matt Madden: http://mattmadden.com/comics/99x/

Yes, if you're insincere, it's patronizing. So it requires empathy and curiosity.

If done right, it's just like the old Dale Carnegie quote:

You can make more friends in two months by becoming interested in other people than you can in two years by trying to get other people interested in you.

Yes! My favorite tactic to get people talking about themselves comes from Paul Ford:

Just ask the other person what they do, and right after they tell you, say: “Wow. That sounds hard.”

Because nearly everyone in the world believes their job to be difficult. I once went to a party and met a very beautiful woman whose job was to help celebrities wear Harry Winston jewelry. I could tell that she was disappointed to be introduced to this rumpled giant in an off-brand shirt, but when I told her that her job sounded difficult to me she brightened and spoke for 30 straight minutes about sapphires and Jessica Simpson. She kept touching me as she talked. I forgave her for that. I didn’t reveal a single detail about myself, including my name. Eventually someone pulled me back into the party. The celebrity jewelry coordinator smiled and grabbed my hand and said, “I like you!” She seemed so relieved to have unburdened herself. I counted it as a great accomplishment. Maybe a hundred times since I’ve said, “wow, that sounds hard” to a stranger, always to great effect. I stay home with my kids and have no life left to me, so take this party trick, my gift to you.

https://medium.com/message/how-to-be-polite-9bf1e69e888c

- Advertorials. When it does work well: The Onion?

https://contently.com/strategist/2016/05/23/dont-make-a-comm...

My bad, thank you for the correction!

This is what I'm most excited about:

"LOOPY lets you have a conversation with simulations! You can go from thinking in systems, to talking in systems."

Thread: https://twitter.com/worrydream/status/808399253928218624

I've also found these to be helpful:

The Language and Grammar of Mathematics, from The Princeton Companion to Mathematics, by James Gowers: http://press.princeton.edu/chapters/gowers/gowers_I_2.pdf

Reading Mathematics, by John Hamal Hubbard: http://www.math.cornell.edu/~hubbard/readingmath.pdf

Additional resources I'm finding helpful:

https://betterexplained.com/articles/an-interactive-guide-to...

http://jackschaedler.github.io/circles-sines-signals/

https://books.google.com/books/about/Who_is_Fourier.html?id=...

"Audio is assumed to be spoken: not suitable for song captioning"

Can anyone recommend alternative approaches for music lyrics alignment?

So your concern isn't just with anecdotes, but with observational data in general?

I agree that experimental data is preferred, but sometimes observational data is all we've got, right?

"You may have heard the phrase the plural of anecdote is not data. It turns out that this is a misquote. The original aphorism, by the political scientist Ray Wolfinger, was just the opposite: The plural of anecdote is data."

http://blog.danwin.com/don-t-forget-the-plural-of-anecdote-i...

"Machine Learning Algorithms that Matter, like Machine Translation, Spam Filter or Cat Generation, will find its way inside a Web Browser." https://twitter.com/hardmaru/status/834607972923748353

Looks like Chapter 15 in the Causal Inference Book agrees with you:

“Outcome regression and various versions of propensity score analyses are the most commonly used parametric methods for causal inference. You may rightly wonder why it took us so long to include a chapter that discusses these methods. So far we have described IP weighting, the g-formula, and g-estimation–the g-methods. Presenting the most commonly used methods after the least commonly used ones seems an odd choice on our part. Why didn’t we start with the simpler and widely used methods based on outcome regression and propensity scores? Because these methods do not work in general. More precisely, the simpler outcome regression and propensity score methods–as described in a zillion publications that this chapter cannot possibly summarize–work fine in simpler settings, but these methods are not designed to handle the complexities associated with causal inference for time-varying treatments.”

Adam, I've been using Machine Learning Is Fun Part 1 at work, to introduce non-technical business leaders to supervised machine learning concepts. Thanks for the great series!

Have you seen the new paper, “Human Decisions and Machine Predictions”? http://scholar.harvard.edu/files/sendhil/files/w23180.pdf

I'm wondering if their methodology is reasonable?

From the abstract: “Millions of times each year, judges must decide where defendants will await trial—at home or in jail. By law, this decision hinges on the judge’s prediction of what the defendant would do if released. … Yet comparing the algorithm to the judge proves complicated. … We only observe crime outcomes for released defendants, not for those judges detained. This makes it hard to evaluate counterfactual decision rules based on algorithmic predictions. … We deal with these problems using different econometric strategies, such as quasi-random assignment of cases to judges. … A policy simulation shows crime can be reduced by up to 24.8% with no change in jailing rates, or jail populations can be reduced by 42.0% with no increase in crime rates. Moreover, we see reductions in all categories of crime, including violent ones. Importantly, such gains can be had while also significantly reducing the percentage of African-Americans and Hispanics in jail. … While machine learning can be valuable, realizing this value requires integrating these tools into an economic framework: being clear about the link between predictions and decisions; specifying the scope of payoff functions; and constructing unbiased decision counterfactuals.”

> I still find most of the mechanics rather shady.

Interesting, can you expand on this? I have no experience with causal inference and would like to learn more. Thanks!

Code: https://github.com/ctford/african-polyphony-and-polyrhythm

15 years later: "Police believe Hogue, a notorious con man who once posed as a high school student when he was 26 years old, built a cabin above the Shadow Mountain Condominiums atop Aspen Street near Lift 1A, and lived there for an undetermined amount of time, though it could have been a year and a half to two years, Detective Jeff Fain has said."

http://www.aspentimes.com/news/con-man-arrested-at-pitkin-co...

Style transfer and generative adversarial networks have worked well with images. How can these be applied to music?

Along those lines, here's a quick read about how "life is a bandit problem."

"When you have more information about what works and what doesn’t, you shift to spending the majority of your time pulling the best lever (exploitation), but you keep exploring the other options in case your current best option isn’t the very best that exists. Here’s the thing: the exploration phase never stops."

https://joshkaufman.net/explore-exploit/

“The answer to anyone who talks about the surplus population is to ask him whether he is the surplus population; or if he is not, how he knows he is not.” —G.K. Chesterton

Also relevant? "Why Some Men Pretend to Work 80-Hour Weeks" https://hbr.org/2015/04/why-some-men-pretend-to-work-80-hour...

How to read research papers & math notation http://blizzard.cs.uwaterloo.ca/keshav/home/Papers/data/07/p... http://press.princeton.edu/chapters/gowers/gowers_I_2.pdf http://www.math.cornell.edu/~hubbard/readingmath.pdf

Fourier transform https://betterexplained.com/articles/an-interactive-guide-to... http://jackschaedler.github.io/circles-sines-signals/ https://books.google.com/books/about/Who_is_Fourier.html?id=...

Music information retrieval http://musicinformationretrieval.com/ https://www.audiolabs-erlangen.de/fau/professor/mueller/book...

I'm very excited for this in Part 2:

"A key teaching goal for us is that you come away from the course feeling much more comfortable reading, understanding, and implementing research papers. We’ll be sharing some simple tricks that make it much easier to quickly scan and get the key insights from a paper."

My interest is musical style transfer. I'd like to replicate these examples from Sony Computer Science Lab-Paris: http://www.flow-machines.com/odetojoy/

They've published papers, but not code (except for DeepBach).