Hi all, I’m building up a code that evaluates the energy of a lot of spin states, and many of those states have the same energy. Typically about 2^20 different states, but only a few thousand of them have different energies. I want to know how many of these energies are different, so I evaluate all…

You can use “trunc”: data = rand(10) unique(trunc.(data, digits = 4))

You can pass a function to unique: unique(x -> round(x, digits=0), randn(10)) Perhaps this will be unique(round(_, digits=0), randn(10)) soon…

Maybe related to: [image] How to make a Set of real values based on rtol? General Usage Lets say you wanted to make a Set of real numbers: Set( [ 1.0, 1+eps(), 2 ] ) Where for all practical purposes, this should reduce to Set([1.0,2.0]) Is there an existin…

Someone periodically asks this, but note that: [image] Is it safe to compare rounded float values for equality? General Usage Any bucketing of floating-point numbers into more than one bucket will have the property that there are values which only differ in the last bit y…

Truncation before unique unfortunately induces a discontinuity if some energy levels are close to the truncation boundary. An easy variant is the following (if your energy is one-dimensional and you know the discretization parameter apriori): function discretize(energy_levels, epsilon) res = Vector…

[image] improbable22: You can pass a function to unique: unique(x -> round(x, digits=0), randn(10)) I would normally use sigdigits=n instead of digits. Using digits is like an absolute tolerance, and is sensitive to the overall scaling of the data, whereas sigdigits is a scale-invariant rel…

Oh, thanks for the replies and the different proposals… I will try them at once :slight_smile: Best, Ferran.

The problem with trunc and round is they work in a given direction. for example if you trunc the last digit of 2.00 and 2.03 you will get 2.0 and 2.0 and they will be equal. But 2.49 and 2.51 will be 2.4 and 2.5 and will be different. One solution could be to sort all numbers, calculate their diff…

I would do something roughly like this: sort and then diff all the values, drop the zeros from the differences, establish a cutoff \Delta from the rest (eg using a quantile, I would plot first), consecutive sorted values within a distance of \Delta go in the same bin, with a sanity check that the…

[image] Juan: One solution could be to sort all numbers, calculate their differences and then apply then decide how the numbers to pick … [image] Tamas_Papp: I would do something roughly like this: That’s exactly what my proposal does. The general approach (higher dimensional “energy”…

Unique() to a certain tolerance?

General Usage

djsegal April 10, 2019, 5:13pm 4

Maybe related to:

Topic		Replies	Views
The test isequal(+0.0,-0.0) returns FALSE Numerics isequal	16	1037	June 2, 2020
How to make a Set of real values based on rtol? General Usage question	2	910	August 6, 2018
Questions related to -0.0 and 0.0 General Usage question	9	577	October 21, 2021
Possible bug in unique/Set Internals & Design faq	49	4757	June 15, 2018
Is it safe to compare rounded float values for equality? General Usage faq	11	9404	February 28, 2019

Unique() to a certain tolerance?

Related topics