Replace values in numpy arrays not working

Question

I have the following problem, and I'd be glad if somebody could explain to me why this is happening.

So, I have this numpy array:

[[ 0.          2.68530063  2.68530063  4.65107712  4.65107712  5.37060126
   2.68530063  4.65107712  4.65107712  5.37060126]
 [ 2.68530063  0.          4.65107712  2.68530063  5.37060126  4.65107712
   4.65107712  2.68530063  5.37060126  4.65107712]
 [ 2.68530063  4.65107712  0.          5.37060126  2.68530063  4.65107713
   4.65107712  7.10463766  7.10463766  8.05590189]
 [ 4.65107712  2.68530063  5.37060126  0.          4.65107713  2.68530063
   7.10463766  4.65107712  8.05590189  7.10463766]
 [ 4.65107712  5.37060126  2.68530063  4.65107713  0.          2.68530063
   7.10463766  8.05590189  9.30215425  9.68198911]
 [ 5.37060126  4.65107712  4.65107713  2.68530063  2.68530063  0.
   8.05590189  7.10463766  9.68198911  9.30215425]
 [ 2.68530063  4.65107712  4.65107712  7.10463766  7.10463766  8.05590189
   0.          5.37060126  2.68530063  4.65107713]
 [ 4.65107712  2.68530063  7.10463766  4.65107712  8.05590189  7.10463766
   5.37060126  0.          4.65107713  2.68530063]
 [ 4.65107712  5.37060126  7.10463766  8.05590189  9.30215425  9.68198911
   2.68530063  4.65107713  0.          2.68530063]
 [ 5.37060126  4.65107712  8.05590189  7.10463766  9.68198911  9.30215425
   4.65107713  2.68530063  2.68530063  0.        ]]

And I want to perform replacements as follows: a) values in array == 0, I want them to be 1. b) values in array == 2.68530063, I want them to be -1, and c) values > 2.68530063 I want them to be 0.

I found that a very compact a nice way of doing it would be this one:

distances[distances==0.00000000]  = 1
distances[distances > 2.68530063] = 0
distances[distances==2.68530063]  = -1

However, my new array output is this:

[[ 1.         -1.          2.68530063  0.          0.          0.
   2.68530063  0.          0.          0.        ]
 [-1.          1.          0.          2.68530063  0.          0.          0.
   2.68530063  0.          0.        ]
 [ 2.68530063  0.          1.          0.          0.          0.          0.
   0.          0.          0.        ]
 [ 0.          2.68530063  0.          1.          0.          0.          0.
   0.          0.          0.        ]
 [ 0.          0.          0.          0.          1.         -1.          0.
   0.          0.          0.        ]
 [ 0.          0.          0.          0.         -1.          1.          0.
   0.          0.          0.        ]
 [ 2.68530063  0.          0.          0.          0.          0.          1.
   0.          0.          0.        ]
 [ 0.          2.68530063  0.          0.          0.          0.          0.
   1.          0.          0.        ]
 [ 0.          0.          0.          0.          0.          0.          0.
   0.          1.         -1.        ]
 [ 0.          0.          0.          0.          0.          0.          0.
   0.         -1.          1.        ]]

My very big doubt here is: why some values == 2.68530063 remains in my numpy array? I really don't get why given that the condition I imposed should be for all values == 2.68530063 (just as it works correctly for the other conditions). And second, what would be then a way of replace correctly a numpy array. I'd like to thank you very much in advance, and happy new year 2014.

user2357112 · Accepted Answer · 2014-01-01 09:56:25Z

3

The array elements that display as 2.68530063 may not be exactly equal to 2.68530063. NumPy truncates array elements for display purposes, and rounding error in the computations that produced your array may have caused elements to have values slightly off from what you'd expect. You can use numpy.isclose to perform comparisons within a tolerance:

distances[numpy.isclose(distances, 0)]  = 1
distances[distances > 2.68530063] = 0
distances[numpy.isclose(distances, 2.68530063)]  = -1

Also, note that the order you perform these assignments in is important. When you compare distances > 2.68530063, the distances values that used to equal 0 will now be 1. That might cause bugs if you're not careful.

answered Jan 1, 2014 at 9:56

user2357112

286k32 gold badges490 silver badges569 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

muammar Over a year ago

I will take a look at numpy.isclose. Thanks for pointing out the importance on the order in which assignments are performed.

ssm · Accepted Answer · 2014-01-01 10:01:20Z

0

Did you try:

distances[distances == 0]  = 1
distances[distances > 2.685300635] = 0
distances[distances > 2.685300625 and distances < 2.685300635]          = -1

?

I have seen this behavior before when trying to use specific values. Basically, 2.68530063 may not be exactly equal to 2.68530063 in memory. It is an approximation to the nearest value represented by the specified number of decimal places ...

Edit: The last check should be:

distances[(distances > 2.685300625) * (distances < 2.685300635)] = -1

I think that last one should work ...

edited Jan 1, 2014 at 10:01

answered Jan 1, 2014 at 9:45

ssm

5,4031 gold badge26 silver badges43 bronze badges

7 Comments

user2357112 Over a year ago

The syntax for the third assignment is wrong. and doesn't work for this.

muammar Over a year ago

@user3135460 I think you are right about this assumption of value being different in memory. I tried also this: Check = distances==0 Check2 = distances>2.68530063 Check3 = distances==2.68530063 if np.all(Check): distances[Check] = 1 if np.all(Check2): distances[Check2] = 0 if np.all(Check3): distances[Check3] = -1 print distances

user2357112 Over a year ago

@muammar: Why the np.all checks?

ssm Over a year ago

sorry about that. shoud have checked it. You can try something aong the lines of: distances[(distances > 2.685300625) * (distances < 2.685300635)] = -1

user2357112 Over a year ago

@muammar: That's not what np.all does, though. np.all(Check1) would only be true if every single element of distances was equal to 0.

|

Collectives™ on Stack Overflow

Replace values in numpy arrays not working

2 Answers 2

1 Comment

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related