oli5679

15 ms·

121.
by oli5679 4y ago
Below is some interesting speed benchmarking. "A series of tests were run to measure the relative performance of SQLite 2.7.6, PostgreSQL 7.1.3, and MySQL 3.23.41. The following are general conclusions drawn from these experiments: SQL
122.
by oli5679 4y ago
I wish them the best of luck in this endeavour!
123.
by oli5679 4y ago
If you treat this as a Bayesian problem, your context and experiences could lead you to specify a prior which has a finite mean, in which case for some values you will have a posterior probability of > 2/3 of it being the larger val
124.
The Scaling Hypothesis (2021)(gwern.net)
97 points by oli5679 4y ago|63 comments
125.
Happiness vs. Social Media Use(nature.com)
2 points by oli5679 5y ago|0 comments
126.
by oli5679 5y ago
How do you avoid adverse selection from traders that have lower latency? E.g if £/$ drops from 1.94 to 1.92 can someone aware of this before you convert £ to $ on your platform?
127.
by oli5679 5y ago
It depends on your domain. If you have a relatively small number of features and your features have relatively stable but not necessarily linear partial dependence, then shap can unpack them pretty well. E.g. value of a house it might not n
128.
by oli5679 5y ago
Did they catch exceptions and then print 'not ok'?
129.
by oli5679 5y ago
This is a really thorough article on the topic [1]. Here is the conclusion. "American antitrust laws date from a time when changes in transportation, communications, and manufacturing technologies generated unprecedented economies of s
130.
by oli5679 5y ago
In most cases monopoly is bad, but with minting currency there is some sense in which it is good for a seniorage rents to exist. Imagine demand for holding USD grows by $20, a central bank can profit from the minting, e.g. if they make a $2
131.
by oli5679 5y ago
If it wasn't for the 6-guess limit and you had a vocabulary, then you could use minimax to try and minimize the maximum number of guesses. E.g. assuming you make the best possible future guesses but your opponent has chosen the least c
132.
Google Python Style Guide(google.github.io)
2 points by oli5679 5y ago|0 comments
133.
by oli5679 5y ago
Huge amounts of analysis and data-recording happen with people juggling spreadsheets, because they are easier for 'business/finance' people to learn how to use them vs. coding and databases. I think there is huge potential fo
134.
by oli5679 5y ago
Ah apologies, I didn't see one yesterday but makes sense to post it on the final day of 2021.
135.
Ask HN: Predictions for 2022?
32 points by oli5679 5y ago|16 comments
136.
by oli5679 5y ago
One thing that is important here is whether the counterfactual is the marginal employee, or nobody. In the sports roster case, a soccer/baseball team would be in trouble if they lost one of their average-talent players if they didn&#x2
137.
Polars: The fastest DataFrame library you’ve never heard of(analyticsvidhya.com)
1 points by oli5679 5y ago|0 comments
138.
by oli5679 5y ago
Cool, I think I got confused by nested threads
139.
by oli5679 5y ago
Would this work? Toss a coin - If heads, say: "I am waiting to see what you say and if you suggested a strategy compatible with a colour to coordiate on, I will follow this coordination, otherwise I will toss a coin and next term will
140.
by oli5679 5y ago
I'm posting the entire distribution and describing two data points because the post would have been very long if I described every country. My aim isn't to cherry pick but simply to describe the distribution of outcomes we current
141.
by oli5679 5y ago
Yeah I think there is lots of nuance about individual datapoints. Does the overall distribution match your mental model? To me it's surprising there aren't at more high spending countries with significantly better outcomes measure
142.
by oli5679 5y ago
Can you think of a country with more equal health expenditure than Saudi Arabia that has much better health outcomes than countries with GDP per capita around $30k?
143.
by oli5679 5y ago
The link is PPP adjusted already
144.
by oli5679 5y ago
Thanks this is great
145.
by oli5679 5y ago
Thanks this is great
146.
by oli5679 5y ago
Yes, there are lots of factors like this when comparing any two countries. USA does have uniquely high expenditure supporting your bloat claim. The point is other countries e.g. France, Switzerland, Canada, UK, there are also not significan
147.
by oli5679 5y ago
If anyone wants some interesting correlational data on life expectancy vs. wealth and health expenditure per capita ourworldindata is a good source. At a country-level the correlational relationships are weaker than many people think. - Cos
148.
by oli5679 5y ago
Here is a 1-liner using pandas. I would be interested if it's possible to do it in one-line using something from itertools. import pandas as pd [{'thing':x, 'count':y} for x,y in pd.DataFrame(mylist).group
149.
by oli5679 5y ago
The interesting thing about this is that in machine learning. - squared error loss --> conditional mean - absolute error loss --> conditional median And this seems to be the same thing with voting.
150.
by oli5679 5y ago
If you want to know why the ratings are centred around 1500 it is simply because this is the value in the original glicko2 paper [1]. It doesn't seem unreasonable for them to implement the algorithm as the designer specified. With rati

More ›