Convert array of indices to 1-hot encoded numpy array

Question

Let s say I have a 1d numpy array a   array  1 0 3    I would like to encode this as a 2D one-hot array b   array   0 1 0 0    1 0 0 0    0 0 0 1     Is there a quick way to do this   Quicker than just looping over a to set elements of b  that is

User · Answer

For 1-hot-encoding     one hot encode pandas get dummies array    For Example  ENJOY CODING

User · Answer

If using tensorflow  there is one hot    import tensorflow as tf import numpy as np  a   np array  1  0  3   depth   4 b   tf one hot a  depth     lt tf Tensor  shape  3  3   dtype float32  numpy    array   0   1   0              1   0   0              0   0   0     dtype float32  gt

User · Answer

I am adding for completion a simple function  using only numpy operators       def probs to onehot output probabilities           argmax indices array   np argmax output probabilities  axis 1          onehot output array   np eye np unique argmax indices array  shape 0   argmax indices array reshape -1           return onehot output array   It takes as input a probability matrix  e g          0 03038822 0 65810204 0 16549407 0 3797123              0 02771272 0 2760752  0 3280924  0 33458805       And it will return       0 1 0 0         0 0 0 1

User · Answer

Here is an example function that I wrote to do this based upon the answers above and my own use case   def label vector to one hot vector vector  one hot size 10               Use to convert a column vector to a  one-hot  matrix      Example          vector    2    0    1           one hot size  3         returns                 0    0    1                   1    0    0                   0    1    0         Parameters          vector  np array   of size  n  1  to be converted         one hot size  int  optional  size of  one-hot  row vector      Returns          np array size  vector size  one hot size   converted to a  one-hot  matrix             squeezed vector   np squeeze vector  axis -1       one hot   np zeros  squeezed vector size  one hot size        one hot np arange squeezed vector size   squeezed vector    1      return one hot  label vector to one hot vector vector   2    0    1    one hot size 3

User · Answer

Here is a function that converts a 1-D vector to a 2-D one-hot array      usr bin env python import numpy as np  def convertToOneHot vector  num classes None               Converts an input 1-D vector of integers into an output     2-D array of one-hot vectors  where an i th input value     of j will set a  1  in the i th row  j th column of the     output array       Example          v   np array  1  0  4           one hot v   convertToOneHot v          print one hot v            0 1 0 0 0            1 0 0 0 0            0 0 0 0 1                assert isinstance vector  np ndarray      assert len vector   gt  0      if num classes is None          num classes   np max vector  1     else          assert num classes  gt  0         assert num classes  gt   np max vector       result   np zeros shape  len vector   num classes       result np arange len vector    vector    1     return result astype int    Below is some example usage    gt  gt  gt  a   np array  1  0  3     gt  gt  gt  convertToOneHot a  array   0  1  0  0           1  0  0  0           0  0  0  1      gt  gt  gt  convertToOneHot a  num classes 10  array   0  1  0  0  0  0  0  0  0  0           1  0  0  0  0  0  0  0  0  0           0  0  0  1  0  0  0  0  0  0

User · Answer

Your array a defines the columns of the nonzero elements in the output array  You need to also define the rows and then use fancy indexing    gt  gt  gt  a   np array  1  0  3    gt  gt  gt  b   np zeros  a size  a max   1    gt  gt  gt  b np arange a size  a    1  gt  gt  gt  b array    0    1    0    0             1    0    0    0             0    0    0    1

User · Answer

You can use the following code for converting into a one-hot vector   let x is the normal class vector having a single column with classes 0 to some number   import numpy as np np eye x max   1  x    if 0 is not a class  then remove  1

User · Answer

Just to elaborate on the excellent answer from K3---rnc  here is a more generic version   def onehottify x  n None  dtype float          1-hot encode x with the max value n  computed from data if n is None          x   np asarray x      n   np max x    1 if n is None else n     return np eye n  dtype dtype  x    Also  here is a quick-and-dirty benchmark of this method and a method from the currently accepted answer by YXD  slightly changed  so that they offer the same API except that the latter works only with 1D ndarrays    def onehottify only 1d x  n None  dtype float       x   np asarray x      n   np max x    1 if n is None else n     b   np zeros  len x   n   dtype dtype      b np arange len x    x    1     return b   The latter method is  35  faster  MacBook Pro 13 2015   but the former is more general    gt  gt  gt  import numpy as np  gt  gt  gt  np random seed 42   gt  gt  gt  a   np random randint 0  9  size  10 000     gt  gt  gt  a array  6  3  7       5  8  6    gt  gt  gt   timeit onehottify a  10  188   s    5 03   s per loop  mean    std  dev  of 7 runs  10000 loops each   gt  gt  gt   timeit onehottify only 1d a  10  139   s    2 78   s per loop  mean    std  dev  of 7 runs  10000 loops each

User · Answer

Such type of encoding are usually part of numpy array  If you are using a numpy array like this    a   np array  1 0 3     then there is very simple way to convert that to 1-hot encoding  out    np arange 4     a   None   astype np float32    That s it

User · Answer

p will be a 2d ndarray  We want to know which value is the highest in a row  to put there 1 and everywhere else 0      clean and easy solution   max elements i   np expand dims np argmax p  axis 1   axis 1  one hot   np zeros p shape  np put along axis one hot  max elements i  1  axis 1

User · Answer

You can use  sklearn preprocessing LabelBinarizer   Example   import sklearn preprocessing a    1 0 3  label binarizer   sklearn preprocessing LabelBinarizer   label binarizer fit range max a  1   b   label binarizer transform a  print   0   format b     output     0 1 0 0    1 0 0 0    0 0 0 1     Amongst other things  you may initialize sklearn preprocessing LabelBinarizer   so that the output of transform is sparse

User · Answer

I recently ran into a problem of same kind and found said solution which turned out to be only satisfying if you have numbers that go within a certain formation  For example if you want to one-hot encode following list   all good list    0 1 2 3 4    go ahead  the posted solutions are already mentioned above  But what if considering this data   problematic list    0 23 12 89 10    If you do it with methods mentioned above  you will likely end up with 90 one-hot columns  This is because all answers include something like n   np max a  1  I found a more generic solution that worked out for me and wanted to share with you   import numpy as np import sklearn sklb   sklearn preprocessing LabelBinarizer   a   np asarray  1 2 44 3 2   n   np unique a  sklb fit n  b   sklb transform a    I hope someone encountered same restrictions on above solutions and this might come in handy

User · Answer

I think the short answer is no  For a more generic case in n dimensions  I came up with this     For 2-dimensional data  4 values a   np array   0  1  2    3  2  1    z   np zeros list a shape     4   z list np indices z shape  -1       a     1   I am wondering if there is a better solution -- I don t like that I have to create those lists in the last two lines  Anyway  I did some measurements with timeit and it seems that the numpy-based  indices arange  and the iterative versions perform about the same

User · Answer

Here s a dimensionality-independent standalone solution   This will convert any N-dimensional array arr of nonnegative integers to a one-hot N 1-dimensional array one hot  where one hot i 1     i N c    1 means arr i 1     i N    c  You can recover the input via np argmax one hot  -1   def expand integer grid arr  n classes                 param arr  N dim array of size i 1       i N      param n classes  C      returns  one-hot N 1 dim array of size i 1       i N  C      rtype  ndarray              one hot   np zeros arr shape    n classes        axes ranges    range arr shape i   for i in range arr ndim       flat grids      ravel   for   in np meshgrid  axes ranges  indexing  ij        one hot flat grids    arr ravel       1     assert  one hot sum -1     1  all        assert np allclose np argmax one hot  -1   arr       return one hot

User · Answer

Here is what I find useful   def one hot a  num classes     return np squeeze np eye num classes  a reshape -1      Here num classes stands for number of classes you have  So if you have a vector with shape of  10000   this function transforms it to  10000 C   Note that a is zero-indexed  i e  one hot np array  0  1    2  will give   1  0    0  1     Exactly what you wanted to have I believe   PS  the source is Sequence models - deeplearning ai

User · Answer

In case you are using keras  there is a built in utility for that   from keras utils np utils import to categorical     categorical labels   to categorical int labels  num classes 3    And it does pretty much the same as  YXD s answer  see source-code

User · Answer

Use the following code  It works best   def one hot encode x           argument         - x  a list of labels     return         - one hot encoding matrix  number of labels  number of class      encoded   np zeros  len x   10    for idx  val in enumerate x       encoded idx  val    1  return encoded   Found it here P S You don t need to go into the link

User · Answer

gt  gt  gt  values    1  0  3   gt  gt  gt  n values   np max values    1  gt  gt  gt  np eye n values  values  array    0    1    0    0             1    0    0    0             0    0    0    1

User · Answer

Using a Neuraxle pipeline step    Set up your example   import numpy as np a   np array  1 0 3   b   np array   0 1 0 0    1 0 0 0    0 0 0 1       Do the actual conversion   from neuraxle steps numpy import OneHotEncoder encoder   OneHotEncoder nb columns 4  b pred   encoder transform a     Assert it works   assert b pred    b   Link to documentation  neuraxle steps numpy OneHotEncoder

User · Answer

You can also use eye function of numpy   numpy eye number of classes  vector containing the labels

[python] Convert array of indices to 1-hot encoded numpy array

Examples related to python

Examples related to numpy

Examples related to machine-learning

Examples related to numpy-ndarray

Examples related to one-hot-encoding