numpy/numpy

Numpy main repository

Eric Wieser

@eric-wieser

@pearu: Not yet, but it was proposed on the mailing list, and the main sticking point was difficulty of implementation

Peng Yu

@yupbank

hey channel, i’m trying to add a cumulative median/ stream median to numpy array, but the whole libary is very well abstracted, can anyone point me a direction ? numpy/numpy#14258

Jody Klymak

@jklymak

@yupbank I'd look for array_cumsum in methods.c, and track the logic through there. Looks like there is a generic accumulation function. Maybe it can take "median" as an operation?

Peng Yu

@yupbank

Array_cumsum is a ufunc of add. i have checked accumulation function which doesn’t quite fit, the implementation of cumulative median is more like a sorting algorithm, since i need to maintain a data structure of all the previous data in memory

add or general accumulation doesn’t support that easily

Vishesh Mangla

@XtremeGood

hi someone here?

what's the error in this ?

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-233-6296b41d2324> in <module>
----> 1 class Matrices:
      2 
      3 
      4     def __init__(self, N=10):
      5 

<ipython-input-233-6296b41d2324> in Matrices()
     27 
     28 
---> 29     self.A = sp.csc_matrix(diags([       [self.h/6]*(self.N),   \
     30                                 [self.h/3, *[2*self.h/3]*(self.N-1), self.h/3], \
     31                                      [self.h/6]*(self.N)],               [1, 0, -1] ), dtype=np.float32)

NameError: name 'self' is not defined

Matti Picus

@mattip

seems a missing indent, line 29 should be indented to be inside the __init__ function

Vishesh Mangla

@XtremeGood

can someone explain the functioning of scipy's RK45 function?

Andrew Schaub

@drewschaub

what do you want to know about RK45?

Vishesh Mangla

@XtremeGood

time step

I want to compute value at t=1

starting from t=0

but step is too slow

Andrew Schaub

@drewschaub

have you played around with max_step, atol or rtol?

Vishesh Mangla

@XtremeGood

maxstep yes

but what is atol

rtol?

also I have two extra time independent initial conditions

Vishesh Mangla

@XtremeGood

I basically have initial value as numpy column array

and each time I want the 0, -1 entry to be according to init conditins ie., 0, 0

Andrew Schaub

@drewschaub

how slow is it compared to another method like rk23?

Vishesh Mangla

@XtremeGood

havnt tested but for 6 time steps it takes around a minute

and that time step should be h**2 always where h=1/N and N is lenth of y

y = initial value

Vishesh Mangla

@XtremeGood

hi someone here

how to dump numpy to pickle?

I have an object class of sparse matrices and lambdas etc . How do I pickle that thing

swamyym

@swamyym

import pickle ; pickle.dump(obj, file)

Vishesh Mangla

@XtremeGood

can someone help me in in rk45?

pickle is not important at the moment

Vishesh Mangla

@XtremeGood

hi someone here?

I have written a program to solve the heat eqn using RK45 of scipy using the finite element method

but increasing the number of basis functions the error increases

its counterintuitive

plz help

I m using hat functions homogenous equation

Loïc Estève

@lesteve

@XtremeGood asking precise and focused questions containing code snippets is very likely to improve the quality of the feed-back you get. I would suggest that you read the following resources: https://opensource.guide/how-to-contribute/#communicating-effectively, https://stackoverflow.com/help/how-to-ask and https://stackoverflow.com/help/minimal-reproducible-example.

Vishesh Mangla

@XtremeGood

heateqn.ipynb

see the rk45 class there

Andreas Mueller

@amueller

silly question but is there a reason that 1.17 is not on conda main?

Vishesh Mangla

@XtremeGood

heateqn.ipynb

why is this code taking too much time?

I 'm using sparse matrices.

csr to be more specific

Vishesh Mangla

@XtremeGood

how to cache np.exp
?

Where communities thrive

People

Repo info

Activity