chas

5 ms·

1.
by chas 2mo ago
Chaos theory does provide us with a tool here: The Lyapunov exponent[1] and the related Lyapunov time[2]. These allow you to characterize how far into the future you can expect to be able to predict the behavior of a dynamical system. To Pr
2.
by chas 3y ago
The major ML conferences all have pretty tight page limits, so more expository sentences usually get cut. This also means that papers usually only explain how their work is different from previous work, so they assume you are familiar with
3.
by chas 3y ago
I think it’s worthwhile to distinguish between throughput and latency for these sorts of discussions, rather than just talking about performance since scratchpads are usually better for latency (even best-case latency) and caches are usuall
4.
by chas 3y ago
I think it’s important to note that you are assuming a gradual increase in activity over time as well as adequate food and rest between sessions. In the limit case things like rhabdomyolysis definitely exist and in the less-extreme case, yo
5.
by chas 4y ago
He developed ME/CFS in 2018 without a known cause (to the best of my knowledge).
6.
by chas 4y ago
Jax is a great tool, but it’s really best for training and experimentation. The transformations outlined in this post (amongst others) make it easy to turn simple and straightforward code into high performance parallel code. While this is c
7.
by chas 4y ago
It really depends on what you want in terms of implementation. On one hand, a big chunk of digital logic is about implementing large, high-speed state machines using circuits. On the other hand, regular expressions and things like lexers ar
8.
by chas 4y ago
The combination of lazy evaluation and state mutation/side effects can be pretty difficult to reason about. For example, if you have a function that changes a global variable as a part of a lazy computation, once that function could ha
9.
by chas 4y ago
Funny enough, monoids/semigroups are one of my favorite parts of Haskell and I really miss having an explicit abstraction for them in other languages. A lot of the code I write is looping over all of the values of a data structure to e
10.
by chas 4y ago
But lenses as functional references are very common in the Haskell world, whereas the only place I've heard of anamorphisms (unfolds) being referred to as lenses is in this paper. It also confused me quite a bit when I first ran into i
11.
by chas 4y ago
While the first author of this paper went on to be heavily involved in Microsoft's Reactive Extensions (among many other things), I think it's better to think of it as making common recursive patterns explicit and introducing abs
12.
by chas 4y ago
This is a great paper, but I found it to be very challenging to read when I first ran into it, even though I had some experience programming in Haskell. As a one sentence pitch, this paper is to recursion what if and while are to goto. For
13.
by chas 4y ago
Where are the uses of the calculus of variations with neural nets?
14.
by chas 4y ago
This isn’t an accurate description of the possible side-effects. Anti-depressants can cause meaningful side-effects including changes to appetite, difficulty regulating body temperature, nausea, changes to sleep, and akathisia. They are als
15.
by chas 4y ago
A semi-decidable problem is still pretty bad news from a computational perspective, but I agree that it's not the best example of what I was trying to illustrate. I was aiming for something dramatic and (somewhat) approachable, but end
16.
by chas 4y ago
Excel has the built in MMULT function, but I’m not aware of any built-in support for eigenvalues or eigenvectors. Many people have written such functions though. That said, I would be surprised if Excel spreadsheets were implemented as matr
17.
by chas 4y ago
And for extra magic, since every vector space has a basis, every linear transform between vector spaces with a finite basis can be represented by a finite matrix ( https://en.m.wikipedia.org/wiki/Transformation_matrix ).
18.
by chas 5y ago
For most brushless motors--though not axial flux motors like the one under consideration here--it's much easier to get heat out by passing air axially across the stator. This means that while the surface area is increasing proportional
19.
by chas 5y ago
It sounds like both your work and hobby involve pretty intense coding. If you are coding for 16-18 hours a day, that doesn't leave enough time for sleep, exercise, and socializing. I've found I need a regular amount of all three o
20.
by chas 5y ago
Heat dissipation mostly. Square-cubed laws are brutal for getting the heat out.
21.
by chas 5y ago
Folks are certainly working on it from many dimensions, but it’s a pretty hard problem since getting ground truth data involves making and testing materials, which is a very different problem to automate than training machine learning model
22.
by chas 5y ago
Spherical linear interpolation. ( https://en.m.wikipedia.org/wiki/Slerp )
23.
by chas 5y ago
If you are not familiar, this is part of a very long-running discussion in the philosophy of mathematics: https://plato.stanford.edu/entries/philosophy-mathematics/#F... (For context on the intro to that section,
24.
by chas 5y ago
I also don't have any direct HFT experience, but I do know that HFT firms hire FPGA designers ( https://www.indeed.com/q-Fpga-Engineer-Trading-jobs.html ) which suggests they really care about counting cycles and basical
25.
by chas 5y ago
Monads are a very different sort of abstraction than the kind that are used most of the time in C programming. In particular, they were originally developed in abstract algebra. Algebraic abstractions are super nice and useful for programmi
26.
by chas 5y ago
It is absolutely very dependent on your specific hardware. I would expect dedicated digital signal processor chips to almost always support a high-performance fixed-point multiply-add instruction. In contrast, I would expect chips targeted
27.
by chas 5y ago
The relative performance of float32 vs int32 is heavily dependent on the specific operations you care about and what hardware resources (i.e. area, power, std cells) you have available. While floating point numbers can be cleanly split into
28.
by chas 5y ago
Integer arithmetic is still simpler to implement in hardware and therefore faster than floating point arithmetic, so it is still heavily used for resource-constrained numerical programs. This shows up in signal processing code for e.g. very
29.
by chas 5y ago
One thing to note: the waves in quantum physics have complex amplitudes[0] (e.g. a real number + an imaginary number), rather than the waves in most synthesizers and analog computers which only have a real amplitude. This means that they ca
30.
by chas 5y ago
While I did computer engineering and not computer science, I frequently come back to how useful of an intellectual foundation my college education was. It's basically a whirlwind tour of humanity's greatest hits in a given domain.

More ›