numpy-discussion

[Numpy-discussion] Handling underflow

[<vi...@id...>]

Is there some way of having calculations which cause underflow
automatically set their result to 0.0?  For example when I take
exp(a), where a is a floating point array.
-- 
Victor S. Miller     | " ... Meanwhile, those of us who can compute can hardly
vi...@id...    | be expected to keep writing papers saying 'I can do the
CCR, Princeton, NJ   | following useless calculation in 2 seconds', and indeed
    08540 USA        | what editor would publish them?"  -- Oliver Atkin

Re: [Numpy-discussion] Handling underflow

[John J. L. <ph...@cs...>]

On 22 Feb 2001, Victor S. Miller wrote:

> Is there some way of having calculations which cause underflow
> automatically set their result to 0.0?  For example when I take
> exp(a), where a is a floating point array.

Not 'automatically', but:

a = whatever()
choose(greater(a, MAX), (a, MAX))
answer = exp(-a)

any good?  This is with a 1D array -- I haven't used higher dimensions
much.


John

[Numpy-discussion] checking identity of arrays?

[John J. L. <ph...@cs...>]

I must be missing something obvious: how does one check if two variables
refer to the same array object?


John

Re: [Numpy-discussion] checking identity of arrays?

[Robert K. <ke...@it...>]

On Fri, 23 Feb 2001, John J. Lee wrote:

> I must be missing something obvious: how does one check if two variables
> refer to the same array object?

Python's id() builtin function?

> John

--
Robert Kern
ke...@ca...

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter

RE: [Numpy-discussion] checking identity of arrays?

[Paul F. D. <pa...@pf...>]

if a is b:
...

-----Original Message-----
From: num...@li...
[mailto:num...@li...]On Behalf Of John
J. Lee
Sent: Friday, February 23, 2001 7:10 AM
To: num...@li...
Subject: [Numpy-discussion] checking identity of arrays?



I must be missing something obvious: how does one check if two variables
refer to the same array object?


John


_______________________________________________
Numpy-discussion mailing list
Num...@li...
http://lists.sourceforge.net/lists/listinfo/numpy-discussion

[Numpy-discussion] How operating arrays in a determined axis?

[Aureli S. F. <Aur...@ip...>]

Hello!

I have a N-dimensional array A and I want to operate in one of the axis (k)
with a 1 dimensional array (for instance,  subtracting an array B of length
k). I have looked for some solutions in the manual and did not found any.

So after testing a lot I found following solution:

A=Numeric.array(a)

A.shape=(a1,a2,...,ak,...,aN)	#where N is any int value

B=Numeric.array(b)
B.shape=(ak,)


A_modified=Numeric.reshape(A, (a1*a2*...*aN,ak)) #ak is not included in the
product

result=[]
for i in A_modified:
	result.append(i - B)


and finally reshaping the result appropriately. But it does not seem
neither elegant nor simple.

Is there any more elegant solution? The key point is: how to operate a N-D
array with a 1D in one determined axis?

Thanks in advance.
Regards,
Aureli




#################################
Aureli Soria Frisch
Fraunhofer IPK
Dept. Pattern Recognition

post: Pascalstr. 8-9, 10587 Berlin, Germany
e-mail:au...@ip...
fon: +49 30 39 00 61 50
fax: +49 30 39 17 517
#################################

Re: [Numpy-discussion] How operating arrays in a determined axis?

[Jon S. <js...@wm...>]

Probably this can help you.

Look in the manual about the ... operator, too.


bash-2.02$ python
Python 1.6b1 (#3, Aug 30 2000, 08:32:49)  [GCC 2.7.2.1] on freebsd3
Copyright (c) Corporation for National Research Initiatives.
Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam.
>>> from Numeric import *
>>> a=array([[2,3],[4,5],[6,7]])
>>> b=array([9,10,11])
>>> a.shape
(3, 2)
>>> b.shape
(3,)
>>> print a+b[:,NewAxis]
[[11 12]
 [14 15]
 [17 18]]
>>>


Jon Saenz.				| Tfno: +34 946012470
Depto. Fisica Aplicada II               | Fax:  +34 944648500
Facultad de Ciencias.   \\ Universidad del Pais Vasco \\
Apdo. 644   \\ 48080 - Bilbao  \\ SPAIN

On Mon, 19 Mar 2001, Aureli Soria Frisch wrote:

> Hello!
> 
> I have a N-dimensional array A and I want to operate in one of the axis (k)
> with a 1 dimensional array (for instance,  subtracting an array B of length
> k). I have looked for some solutions in the manual and did not found any.
> 
> So after testing a lot I found following solution:
> 
> A=Numeric.array(a)
> 
> A.shape=(a1,a2,...,ak,...,aN)	#where N is any int value
> 
> B=Numeric.array(b)
> B.shape=(ak,)
> 
> 
> A_modified=Numeric.reshape(A, (a1*a2*...*aN,ak)) #ak is not included in the
> product
> 
> result=[]
> for i in A_modified:
> 	result.append(i - B)
> 
> 
> and finally reshaping the result appropriately. But it does not seem
> neither elegant nor simple.
> 
> Is there any more elegant solution? The key point is: how to operate a N-D
> array with a 1D in one determined axis?
> 
> Thanks in advance.
> Regards,
> Aureli
> 
> 
> 
> 
> #################################
> Aureli Soria Frisch
> Fraunhofer IPK
> Dept. Pattern Recognition
> 
> post: Pascalstr. 8-9, 10587 Berlin, Germany
> e-mail:au...@ip...
> fon: +49 30 39 00 61 50
> fax: +49 30 39 17 517
> #################################
> 
> 
> 
> _______________________________________________
> Numpy-discussion mailing list
> Num...@li...
> http://lists.sourceforge.net/lists/listinfo/numpy-discussion
> 

Re: [Numpy-discussion] How operating arrays in a determined axis?

[Warren F. <fo...@sl...>]

On Mon, 19 Mar 2001, Aureli Soria Frisch wrote:

> Hello!
> 
> I have a N-dimensional array A and I want to operate in one of the axis (k)
> with a 1 dimensional array (for instance,  subtracting an array B of length
> k). I have looked for some solutions in the manual and did not found any.
> 
> So after testing a lot I found following solution:
> 
> A=Numeric.array(a)
> 
> A.shape=(a1,a2,...,ak,...,aN)	#where N is any int value
> 
> B=Numeric.array(b)
> B.shape=(ak,)


newshape = [1] * len(A.shape)
newshape[k] = A.shape[k]
B.shape = newshape
result = A - B

wbf

Re: [Numpy-discussion] How operating arrays in a determined axis?

[John J. L. <ph...@cs...>]

On Mon, 19 Mar 2001, Aureli Soria Frisch wrote:
[...]
> I have a N-dimensional array A and I want to operate in one of the axis (k)
> with a 1 dimensional array (for instance,  subtracting an array B of length
> k). I have looked for some solutions in the manual and did not found any.
[...]
> Is there any more elegant solution? The key point is: how to operate a N-D
> array with a 1D in one determined axis?

I'm not sure I understand exactly what you want.  Are you sure you don't
want to do A[n1,n2,...,:,...,nN] instead?  (the '...'s are just an
indication of where the extra numbers would go, not the literal Numeric
syntax. If so, you can use

s = slice(None)
# n is your list [n1, n2, n3, ..., n(k-1), n(k+1), ...,nN]
n.insert(k, s)
print "answer is", A[n] - B
print slice.__doc__

(possibly there is an off-by-one error in the above, but you get the idea)


John

Re: [Numpy-discussion] How operating arrays in a determined axis?

[Aureli S. F. <Aur...@ip...>]

Thanks to all answers in spite of the not so understandable question.

I have solved my problem NewAxis seems tome the simplest and most elegant
way to solve the problem. Maybe a name like 'VirtualAxis' would clarify
more the question.

Regards

>On Mon, 19 Mar 2001, Aureli Soria Frisch wrote:
>[...]
>> I have a N-dimensional array A and I want to operate in one of the axis (k)
>> with a 1 dimensional array (for instance,  subtracting an array B of length
>> k). I have looked for some solutions in the manual and did not found any.
>[...]
>> Is there any more elegant solution? The key point is: how to operate a N-D
>> array with a 1D in one determined axis?

#################################
Aureli Soria Frisch
Fraunhofer IPK
Dept. Pattern Recognition

post: Pascalstr. 8-9, 10587 Berlin, Germany
e-mail:au...@ip...
fon: +49 30 39 00 61 50
fax: +49 30 39 17 517
#################################

Re: [Numpy-discussion] How operating arrays in a determined axis?

[Konrad H. <hi...@cn...>]

> I have a N-dimensional array A and I want to operate in one of the axis (k)
> with a 1 dimensional array (for instance,  subtracting an array B of length
> k). I have looked for some solutions in the manual and did not found any.

I am not sure that I understand your problem exactly, but here's the solution
to what I think your problem is ;-)

  from Scientific.indexing import index_expression
  import Numeric

  indices = index_expression[::] + \
	    (len(A.shape)-k-1)*index_expression[Numeric.NewAxis]
  result = A-B[indices]

This uses a module from ScientificPython which provides syntactic sugar
for indexing. In fact, the module is so simple that I include it here:

---------------------------------------------------------------------------
# A nicer way to build up index tuples for arrays.
#
# You can do all this with slice() plus a few special objects,
# but there's a lot to remember. This version is simpler because
# it uses the standard array indexing syntax.
#
# Written by Konrad Hinsen <hi...@cn...>
# last revision: 1999-7-23
#

"""This module provides a convenient method for constructing
array indices algorithmically. It provides one importable object,
'index_expression'.

For any index combination, including slicing and axis insertion,
'a[indices]' is the same as 'a[index_expression[indices]]' for any
array 'a'. However, 'index_expression[indices]' can be used anywhere
in Python code and returns a tuple of indexing objects that can be
used in the construction of complex index expressions.

Sole restriction: Slices must be specified in the double-colon
form, i.e. a[::] is allowed, whereas a[:] is not.
"""

class _index_expression_class:

    import sys
    maxint = sys.maxint

    def __getitem__(self, item):
        if type(item) != type(()):
            return (item,)
        else:
            return item

    def __len__(self):
	return self.maxint

    def __getslice__(self, start, stop):
	if stop == self.maxint:
	    stop = None
	return self[start:stop:None]

index_expression = _index_expression_class()
---------------------------------------------------------------------------

Konrad.
-- 
-------------------------------------------------------------------------------
Konrad Hinsen                            | E-Mail: hi...@cn...
Centre de Biophysique Moleculaire (CNRS) | Tel.: +33-2.38.25.56.24
Rue Charles Sadron                       | Fax:  +33-2.38.63.15.17
45071 Orleans Cedex 2                    | Deutsch/Esperanto/English/
France                                   | Nederlands/Francais
-------------------------------------------------------------------------------