Yes, I could have been using THREE.OrthographicCamera(), however, it makes the effect way less cinematic, and most importantly, it makes the depth harder to see (particularly for a scatter plot).
Thanks a lot! That was my goal. It's a trick I learned: if you zoom in from a distance, everything appears flat. The effect is achieved by zooming out while simultaneously moving the camera closer to the subject!
Author here. I'm using Svelte, which is great for interactive applications. For the event handling I'm very influenced by what https://mlu-explain.github.io/ does.
The 3d plot is made with Threejs through the Threlte wrapper. One challenge was animating the 20k points in the 3d plot, which is handled by a custom vertex shader.
That's really nice to hear and it gives motivation to continue! Next article will probably go 3D, but it adds a lot of difficulties. I still need to find out a good small-scale but playful shader project to cover!
Indeed it's a simplification! You can use buffers to keep information from previous states or share state between pixels (for both fragment and compute shaders). Though it can greatly reduce the performance sometimes.
For an intro however, I think it's simpler to assume no memory nor shared state.
I finally found the courage to write and expose myself to the internet. I've always wanted to learn shaders so I thought it would be nice to document my learning and share it with others.
Causality has been the objective of econometrics for many years. I always wonder why people in machine learning overlook it's contributions. They almost never discuss it (even to criticize it) and prefer reinventing the wheel.
> Economists definitely claim to know how things work.
Some do, most don't. Most economists are just trying to publish their work in peer reviewed journals. The one that have time to talk on mass media every weeks are not regular economist. Krugman had some great papers, but he is definitely not considered as a researcher anymore by the profession.
> Allowing completely free trade will be great he said. Prices will be lowered and the economy will run more efficiently. Maybe that’s true, but it also hollowed out the U.S. manufacturing base and consequently the middle class. You don’t see articles like that anymore.
The HOS model, commonly accepted in the trade literature that Krugman knows well, is very clear about the effect of free trade (even if the model could be regared as too simplistic): it lowers the prices and increases total production, however, it creates unequal gains (or even loss) within countries. The standard policy recommandation among trade economists is public polices should be aimed to losers using a fraction of gains from trade.
There may be a huge selection bias in their study: insiders are more likely to publish their game, weather they are bad or good. In comparison, hobbyist have to propose better games in order to have the chance to be visible and then published. This would bias the study into thinking that hobbyist are more innovative than professionals.
As an applied economist, I can assure you that R is wayyy better in terms of statistics than Python. Doing causal models in Python is a real pain because its data science community is mostly focused on machine learning. Even a robust linear model with IV or fixed effect (which is standard in causal models) is really hard to achieve in Python. Of course this argument is about the libraries and note the core language, but it's a strong argument to stay with R. This argument is also valid outside academics, as more and more data scientists try to adress causality.