antipaul

58 ms·

331.
by antipaul 6y ago
That seems like a different topic than privacy’s relationship to their business model...
332.
by antipaul 6y ago
Totally. Often, I see a copy paste of that canonical regularization figure in TL,DR 1. which I’ve never seen adequately explained. Until this post. Edit: Ok and now I see what you did there ;)
333.
by antipaul 6y ago
You can’t put Apple and Google in the same bucket when it comes to privacy. Sure, neither is probably perfect, but only one of them is an egregious violator whose business model depends on surveillance capitalism
334.
by antipaul 6y ago
Just to add perspective, one reason why it’s _not_ great is the deluge of poor quality information and therefore potential for bias and “the silent majority of experts”, where you can read a bunch of stuff and feel like you know a lot, when
335.
by antipaul 6y ago
Help me see how this requires a _thorough_ study of Bayesian methods. What you describe seems to be (additive/Laplace) smoothing [1], a fairly basic concept. And I don't see how this is specifically Bayesian. Then, one common crit
336.
by antipaul 6y ago
And to add, a clinical trial can be thought of as just one step to establish evidence for efficacy and safety. Because if you really want to prove that it works and is safe, you have to give it to a lot of people, all of whom have different
337.
by antipaul 6y ago
Just because it's not reported in the Times, doesn't mean it doesn't happen. It actually can happen quite frequently. Trial protocols can pre-specify early stopping criteria: if we see drug A is better than drug B by a factor
338.
by antipaul 6y ago
The saying in our industry is, "clinical trials are the most complicated thing in the world". EDIT: adding details: You have to mobilize large interdisciplinary teams to come up with products that are inherently very hard to inven
339.
by antipaul 6y ago
Are those really the best starts for "Bayesian statistics"? Especially the first 2 are rather the standard "intro to ML textbooks", with a frequentist focus (ISL may even have zero Bayesian stuff - Naive Bayes is not &qu
340.
by antipaul 6y ago
I support this learning method. Having a map of the key concepts, as well as visceral feel for them via code, will keep up the motivation. This would be in contrast to picking up one of the plethora of “ML” textbooks that mostly only descri
341.
by antipaul 6y ago
Can you give a concrete example how a Bayesian concept provides practical help in ML model training or another ML task? Want to whet my appetite for your suggestion.
342.
by antipaul 7y ago
Yep, that's what I mean. Wouldn't it be better for all and everything to have a higher-functioning tools team?
343.
by antipaul 7y ago
I would like to ask _why_ does this nightmare exist, but is that the right question? What is the right question to ask here.
344.
by antipaul 7y ago
Agree with your points on "old company/conservative industry" and "non-engineering culture" I'm at a place that is both, and both are huge pains. On the engineering side, it's a bit different though: techn
345.
by antipaul 7y ago
Totally agree. Non-tech companies that think they need "data science" should instead put same effort into (data) engineering. It's not quite 99% of the effort but close enough ;) Search "data science hierarchy of needs&q
346.
by antipaul 7y ago
Reasons for joining: 70% just needed a job, 20% location (silicon valley - wanted to be in tech ecosystem), 10% was combo of: I was told it's a small, entrepreneurial team with undefined remit (so opportunities to forge own path - turn
347.
by antipaul 7y ago
100% agree with the article. The top misconceptions are spot on. I’m at a big place where data science hype among leadership couldn’t be bigger.
348.
by antipaul 7y ago
There are some differences between stats and data science. At least initially, DS was a lot about machine learning. While those methods may be statistical, it was the computer science field that drove and embraced the ML revolution. Current
349.
by antipaul 7y ago
Thanks all for the input, it helps. I didn’t mean to downplay anyone’s contributions. Just looking at how big companies in tech vs in healthcare are solving this. Of course, smaller players innovate on a different scale. Cheers!
350.
by antipaul 7y ago
This is impressive, compared to what traditional health companies like Pharma are doing. The latter seem mostly to update existing products like faster tests or trying to develop vaccines. Whereas it’s the tech companies which are going a b
351.
by antipaul 7y ago
I wish all papers were structured this way, by default. That is, plenty of good diagrams, clear explanations and intuitions, no unnecessary mathiness.
352.
by antipaul 7y ago
So what's next for Legere? I thought it was he who was supposed to lead the merged Co...
353.
by antipaul 7y ago
In a survey on "scientific discovery", I would have expected more examples than face and image recognition and natural language processing, which are so stale at this point. Healthcare? Physics? Chemistry? Biology? Sociology?
354.
by antipaul 7y ago
I think you can achieve that with 2 spaces at the end
355.
by antipaul 7y ago
“The necessary conceits to sustain bureaucracies”, well put! “Bullshit Jobs”, anyone ;)?
356.
by antipaul 7y ago
Terrific man. Glad to see folks be creative and disciplined against meeting culture. Around here, everyone complains about meetings, yet few do anything about it. Its just easier to go to meetings, instead of thinking hard to reduce others’
357.
XGBoost with Nested Cross-validation (Python code example)(github.com)
1 points by antipaul 7y ago|0 comments
358.
by antipaul 7y ago
Thank goodness it was "aided by machine learning". Otherwise, would anybody be reading this article ;)?
359.
by antipaul 7y ago
This is why McKinsey, others have “Up or out” policy. In theory, it can prevent all this bs. In practice, hopefully it prevents a lot of it
360.
by antipaul 7y ago
Pretty sure you can use PIN, including 6-digit and alphanumeric Or FaceID

More ›