oli5679

34 ms·

301.
by oli5679 7y ago
https://ourworldindata.org/coronavirus-testing-source-data The key metric is tests/confirmed case. When I've looked, this is very predictive of confirmed cases (2 weeks ago) per death.
302.
by oli5679 7y ago
I'm posting with the article title but agree that it's a bit clickbaity. Please change if you can think of something more moderate.
303.
World of Warcraft perfectly predicted our coronavirus panic(wired.co.uk)
1 points by oli5679 7y ago|1 comments
304.
by oli5679 7y ago
https://github.com/CSSEGISandData/COVID-19 Plot confirmed cases and deaths vs. date on a log scale for countries with more than N cases/deaths. You'll see USA following similar pattern to Italy, and other Eur
305.
by oli5679 7y ago
This thread is quite encouraging. Argues that reason why we don't have one for other coronaviruses is because of cost-benefit not high enough, rather than technical hurdles. 20% colds caused by 4 different Coronaviruses, so hard to jus
306.
Neil Ferguson: coronavirus expert who is working on despite symptoms(theguardian.com)
4 points by oli5679 7y ago|0 comments
307.
Bill Gates Covid-19 AMA(reddit.com)
15 points by oli5679 7y ago|0 comments
308.
by oli5679 7y ago
Luckily we make a lot of tonic water!
309.
by oli5679 7y ago
Naive question, does this mean quinine may also help, and is there enough in tonic water to have any clinical effect?
310.
by oli5679 7y ago
The paper is here - https://www.thelancet.com/journals/laninf/article/PIIS1473-3...
311.
UK Government Advisor's Ncov model code(github.com)
1 points by oli5679 7y ago|1 comments
312.
by oli5679 7y ago
I wrote a Selenium bot to unsubscribe from notifications from all my friends. Now I just use Facebook for messaging and events.
313.
by oli5679 7y ago
This is just modelling transfers. A lot of economics focusses on gains from interaction/trade.
314.
by oli5679 7y ago
I also found this summary of the writing process quite interesting, by Tyler Cowen: https://www.writingroutines.com/tyler-cowen/
315.
by oli5679 7y ago
Open source implementation of Microsoft Excel, which lets you write macros using Python and has the plotting, scheduling and report sharing functionality of Tableau.
316.
by oli5679 7y ago
Is anyone familiar with pros/cons vs. AWS ecs? I've used ECS a few times, it's pretty nice. https://aws.amazon.com/ecs/
317.
by oli5679 7y ago
Information is power. But like all power, there are those who want to keep it for themselves. The world's entire scientific and cultural heritage, published over centuries in books and journals, is increasingly being digitized and lock
318.
by oli5679 7y ago
Any way to sort by points? Edit - https://hn.algolia.com/?query=&sort=byPopularity&prefix&page...
319.
How to Speak by Patrick Winston(youtube.com)
2 points by oli5679 7y ago|0 comments
320.
New approaches shed light on the magnitude of sex differences in personality(blogs.scientificamerican.com)
262 points by oli5679 7y ago|328 comments
321.
by oli5679 7y ago
If you aren't wedded to R then pickling sklearn Pipeline and loading in Flask app can be nice. Advantage of this is that data pre-processing can also be included in a sklearn Pipeline. https://scikit-learn.org/stable&#x
322.
by oli5679 7y ago
I didn't see you his comment when I posted but I would strongly recommend PMML-based approach.
323.
by oli5679 7y ago
I'd recommend exporting R model as PMML file, and getting your Java team to interact with Openscoring server. PMML is language agnostic model specification (XML like). Python and R machine learning ecosystem can easily generate these (
324.
by oli5679 7y ago
I really appreciate the philosophy of defining metric, and measuring performance going from simple to complex methods. I've also used Prophet library, and find it works well out of the box.
325.
by oli5679 7y ago
Reminds me of this: https://xkcd.com/327/
326.
by oli5679 7y ago
Sparse decision trees and regression models are easily explainable. In some domains, with sufficient feature engineering you can get reasonable results with these approaches.
327.
by oli5679 7y ago
Statsmodels has a nice implementation of logistic regression more suited for statistical analysis. It has no default regularisation, nice labelled coefficients table, and model evaluation metrics. https://www.statsmodels.org&#x2F
328.
by oli5679 7y ago
Yes that is valid point about the limitation of this approach. There are 3 methods to address it, none of them are perfect: (1) Statistically compare the population of twins vs the population of non-twins. The paper I linked to has more spe
329.
by oli5679 7y ago
There is extensive literature on twin studies that finds effects of roughly this magnitude on a range of health/earnings/educational metrics. 40%-50% genes 40%-50% random/non-shared environments and small impact of shared env
330.
by oli5679 7y ago
If your model can be exported as PMML, this is really nice. Fast, minimalist, battle-tested and with very clean API. When I've tested, it's up to 10x faster than Flask + serialised model object and uses far less CPU resources. Pla

More ›