aifer4

5 ms·

1.
Solving the compute crisis with physics-based ASICs(arxiv.org)
5 points by aifer4 1y ago|2 comments
2.
by aifer4 1y ago
This is a perspective article I wrote with a number of researchers in unconventional computing, about solving problems in AI using physics-based chips. There is also a video of me discussing the paper on a livestream here https:/&#x2F
3.
Thermodynamic Bayesian Inference(arxiv.org)
7 points by aifer4 2y ago|1 comments
4.
by aifer4 2y ago
Latest research from Normal Computing
5.
The Specialness of Groups of 3(maxaifer.wordpress.com)
1 points by aifer4 2y ago|1 comments
6.
by aifer4 2y ago
I've been meaning to start this blog forever, and this is my first post
7.
by aifer4 3y ago
> I don’t know what I may seem to the world, but as to myself, I seem to have been only like a boy playing on the sea shore, and diverting myself in now and then finding a smoother pebble or a prettier shell than ordinary, whilst the gre
8.
by aifer4 3y ago
You're right, that iterative methods may be linear in time with respect to the number of parameters. In this paper, we provide a method which is linear in time with respect to the dimension (d) of the vector (x) we are solving for in t
9.
by aifer4 3y ago
To answer your question about sorting an array: For an array of length n, where each element takes one of m possible values, there are n^m possible arrays. But there are only O(n^m/n!) possible sorted arrays, which could be crudely app
10.
by aifer4 3y ago
The Optical Fourier Transform (OFT) in [2] is a way to compute the Discrete Fourier Transform (DFT), and is an alternative to the Fast Fourier Transform (FFT). The OFT basically does a matrix-vector multiplication, where the matrix has a sp
11.
by aifer4 3y ago
The relationship with Hopfield networks sounds fascinating, would love to discuss further. As you mentioned, there is a connection to annealing in that we are encoding the solution to our problem in the minimization of a physical system&#x2
12.
by aifer4 3y ago
Had a quick read of [3]. This work comments on three contributions to the error in an analog scheme to solve linear systems via an ODE that is encoded in the circuit dynamics, although the specific ODE being solved is not given in the paper
13.
by aifer4 3y ago
We do not (yet?). Here is a simulation I made of similar hardware demonstrating the equilibration step of the algorithm https://app.normalcomputing.ai/composer/playground
14.
by aifer4 3y ago
https://app.normalcomputing.ai/composer
15.
by aifer4 3y ago
An interesting feature of this approach is that the proposed hardware doesn't rely on non-linear elements, memristors, or even active elements (besides an optional noise source). It is simply a passive network of oscillators with a DC
16.
by aifer4 3y ago
You mentioned that such problems may be solved by solving a linear approximation of the non-linear problem (and I am in no way an expert in non-linear optimization). To the extent that the bottleneck in that approach is solving the resultin
17.
by aifer4 3y ago
One way to think about these methods is that we are essentially implementing a Monte-Carlo algorithm physically, where on each "iteration" there is a matrix-vector multiplication. The physical system does this matrix-vector multip
18.
by aifer4 3y ago
These results probably would not hold in the same form for a quantum system. By a quantum system, I mean a system where the decoherence time is on the order of the other timescales present in the system (e.g. the correlation time). In fact
19.
by aifer4 3y ago
Straying a bit off topic, but I think one of the more sensible approaches to information-energy equivalence is a thermodynamic engine with an information reservoir (in addition to the heat and work reservoirs normally considered). https:&#
20.
by aifer4 3y ago
Definitely. Landauer's principle gives a lower bound on the amount of energy a computation requires, which is k_B T ln(2) times the number of bits erased in the process, the decrease in Shannon entropy. Our energy cost analysis is not
21.
by aifer4 3y ago
This is an important observation, and is one of the reasons we included an energy-time tradeoff analysis in the paper. To our knowledge, this is the first result where the product energy * time has been shown to scale with dimension for sol
22.
by aifer4 3y ago
Good question. There are different ways that errors come into analog computations, including thermal noise, measurement imprecision, and imprecision of the device's physical parameters. This work addresses thermal noise (which is alway
23.
Thermodynamic Linear Algebra(arxiv.org)
234 points by aifer4 3y ago|55 comments
24.
by aifer4 3y ago
Latest research from https://normalcomputing.ai/ . Feedback welcome!
25.
by aifer4 4y ago
This is a great technical demonstration of how to visualize a circuit board. But I would rather walk over hot coals than try to actually design a circuit using this interface. Circuits can be represented very nicely in two dimensions, which
26.
by aifer4 6y ago
I wonder if anyone’s written a parser for this and applied it to a large number of songs. I would be curious to see how often each rule is used. On a related note, I would be interested to know whether the grammar is ambiguous.
27.
by aifer4 6y ago
The definition of violence: "behavior involving physical force intended to hurt, damage, or kill someone or something." Unless we're using words in the humpty dumpty sense of "it means just what I choose it to mean—neith