gcmac

I don't understand why they don't just make the "reciprocity" argument. The fact that we allow Chinese social networks in the US while they block ours doesn't pass the smell test. To me, this line of argument is easier to rally around as it's more business focused and avoids free speech/national security debates which only serve to muddy the waters

I wonder if their A/B test showed that this was somehow "solving" the "spam of spam reports" problem. It would be fairy naive but I could envision a world where they tested this out, saw fewer (but more detailed) spam reports and concluded that the prior distribution of spam reports was itself not reliable (e.g overcounted the problem).

This is obviously self-serving but seems like it may at least be a "consistent" world view where they aren't totally cynical.

Reminds me of this excerpt from an Obama interview:

> When problems reached him in the White House, he said, it was because they were unsolvable. He generally was being asked to choose between two bad options. “By definition, if it was an easily solvable problem, or even a modestly difficult but solvable problem, it would not reach me, because, by definition, somebody else would have solved it,” he said. “So the only decisions that came were the ones that were horrible and that didn’t have a good solution."

Not sure how heavily used it is, but FWIW Postmates does offer a $10/month subscription service for reduced fees/free delivery like most other food delivery services do

I’d say it’s more likely they have super advanced/clever ways of doing the latter. The algorithm could be a simple dot product and the result could be great or terrible depending on how good the feature extraction is.

Pulling useful features out of videos is no small task. The fact that everyone raves about how good the recommendations are indicates to me that this is where their innovation lies.

Again definitely not a lawyer but I’d imagine advertising an ETF or selling a subscription to an email list based on these trades (both ideas from this thread) would be pretty clear violations.

Looks like the data may be coming from https://efdsearch.senate.gov/search/home/

IANAL but I’d tread carefully given the language around not being used for commercial purposes

This is a pretty cool video about a couple that made the MLB schedule for years https://youtu.be/yT0CMOGKKhU

Having been through this recently I certainly empathize with feeling like you're doing something wrong when looking for a new job. What got me through it was realizing that if the company was looking to replace me they would do the exact same thing - it's just how business works.

That being said, as difficult as it is to continue to give 100% leaving well is very important for building references that could help you in the future.

Also worth reading his book "Disrupted" - the book is every bit as funny and relevant as Silicon Valley

Thanks for posting - totally worth the read to learn there's a pd.read_clipboard() function.

Not Jeremy (obviously) but having taken both versions of the course I would definitely recommend jumping into the 2018 version. The core content is the same (image recognition, recommender systems, NLP) but the 2018 version uses more cutting edge tools and has a tighter feel (as a v2 of anything usually does).

Go write a couple hundred lines of R (please don't actually do this R is atrocious imo) and then you'll understand why '=' is used instead of '<-' - because it's a pain in the ass, even with hotkeys

Very informative and well written article about KNN classification. However, as a data scientist it always pains me to see the iris data set being used. It is linearly separable and gives no indication of whether or not you actually want to use the given methodology on your problem since almost every technique can achieve these results on this data. I'd recommend using something from kaggle or even the UCI repository to make these types of articles even more useful!

I disagree with the analysis of this article. In a typical machine learning process, the response variable stays the same (at a distributional level) but you cycle through candidate models. So regardless of whatever the class distributions are, a higher AUC score indicates a better model.

It might be true that the classifier performance is worse on an imbalanced data set (with the same AUC score) than a balanced one but that just reflects the fact that classifiers are harder to build for imbalanced data.

I used to laugh at stuff like this, then at orientation the place I was working with said "The best way to show your personality is through your socks." I noped right out of there.