I just discovered this field; I needed an algorithm to compute a low rank approximation of a large covariance matrix without first building the matrix. I'm using algorithm 16 in this paper which is simple but very good!
110 karma · joined August 18, 2015
With Jax you can write native for loops that can also be jitted (I imagine you can also do this in numba?); this can then be really fast. Though in that case you would have to write the optimisation algorithm yourself which is not always practical!
Another big speedup in Jax is due to vmap/pmap, which allow to vectorise/parallelise computation. For example you can build a massive gram matrix really quickly using vmap.
Another point: Jax can also run on GPU (like numba :) ) without having to rewrite anything.
The main selling point of the library is automatic differential and compilation to XLA, but I've been using it even when I don't need gradients, as it's really fast (due to compilation). I also really like the random number generator as it's very good for reproducibility.
I've played around with Julia in the past and really liked it, but in terms speed Jax has pretty much solved that problem for me
It feels really similar to modular arithmetic: -3 = 7 (modulo 10)
It's a system of ODEs that describes the dynamics of each driver. So the inputs are: the parameters of the model and the inflow of cars in the section of road
I’ve recently started using JAX and I’m really into it! It’s really fast even on my laptop (cpu) and can run even faster on GPU. I’m now using jax even for random scripts as well as I prefer the random number generator to numpy’s one. This is because you pass the random keys explicitly to all random functions so everything is reproducible.
I really like programming side the research; especially trying to get algorithms to be fast! So Jax is really great for that
I’ve started it recently and I’m writing about computational statistics and programming-related stuff. The idea is to write things that I would have liked to read about (I do computational stats, and also like web development).
So far I’ve written about a goodness of fit test for MCMC code, a D3 visualisation for the gypsy jazz scene around the world, and a classic “how to implement natural numbers from scratch” in OCaml.
For me: what helped was easing into it: I would only work a bit from home for an hour or 2 in the morning before going into the office (it was during my PhD so no one was monitoring my time).
When I got more comfortable working from home I started doing entire days, but not the entire week. Eventually, I started working only from home. This way meant that I could set up good habits (they'll be different for everyone) before doing it all the time.
It shows which players gigged together, and colours the players based on the country they live in. It's a wiki, so anyone can add themselves or other players, as well as add a link to a YouTube video to promote their music.
A bunch of people (Gypsy jazz a pretty small world!) added themselves (there are over 200 players on it now) then it died down.
I also wrote a blog post about the design process: https://www.jeremiecoullon.com/2019/11/27/djangoverse/
It was fun and meant I got to learn a bit of React and D3 !
How many email addresses did you try for each CEO ? I'm guessing the most common one is firstname.lastname@company.com ?