# onehotdecode

Decode probability vectors into class labels

## Description

decodes each probability vector in `A`

= onehotdecode(`B`

,`classes`

,`featureDim`

)`B`

to the most probable class label from the labels specified by `classes`

. `featureDim`

specifies the dimension along which the probability vectors are defined. The function decodes the probability vectors into class labels by matching the position of the highest value in the vector with the class label in the corresponding position in `classes`

. Each probability vector in `A`

is replaced with the value of `classes`

that corresponds to the highest value in the probability vector.

## Examples

### One-Hot Encode and Decode Labels

Encode a set of labels into probability vectors, and then decode them back into labels.

Create a vector of categorical labels specifying color types.

colorsOriginal = ["red","blue","red","green","yellow","blue"]; colorsOriginal = categorical(colorsOriginal)

`colorsOriginal = `*1x6 categorical*
red blue red green yellow blue

Determine the classes in the categorical vector.

classes = categories(colorsOriginal)

`classes = `*4x1 cell*
{'blue' }
{'green' }
{'red' }
{'yellow'}

One-hot encode the labels into probability vectors by using the `onehotencode`

function. Encode the labels into the first dimension, so that each row corresponds to a class and each column corresponds to a probability vector.

colorsEncoded = onehotencode(colorsOriginal,1)

`colorsEncoded = `*4×6*
0 1 0 0 0 1
0 0 0 1 0 0
1 0 1 0 0 0
0 0 0 0 1 0

Decode the probability vectors by using the `onehotdecode`

function.

colorsDecoded = onehotdecode(colorsEncoded,classes,1)

`colorsDecoded = `*1x6 categorical*
red blue red green yellow blue

The decoded labels match the original labels.

### One-Hot Decode Dummy Variables

Create dummy variables, and then decode them back into the original data.

Create a column vector of categorical data specifying color types.

colorsOriginal = ["red";"blue";"red";"green";"yellow";"blue"]; colorsOriginal = categorical(colorsOriginal)

`colorsOriginal = `*6x1 categorical*
red
blue
red
green
yellow
blue

Determine the classes in the categorical vector.

classes = categories(colorsOriginal);

Create dummy variables for each color type by using the `dummyvar`

function.

dummyColors = dummyvar(colorsOriginal)

`dummyColors = `*6×4*
0 0 1 0
1 0 0 0
0 0 1 0
0 1 0 0
0 0 0 1
1 0 0 0

Decode the dummy variables in the second dimension by using the `onehotdecode`

function.

colorsDecoded = onehotdecode(dummyColors,classes,2)

`colorsDecoded = `*6x1 categorical*
red
blue
red
green
yellow
blue

The decoded variables match the original color types.

### Decode Probability Vectors into Most Probable Classes

Decode a set of probability vectors into the most probable class for each observation.

Create a set of 10 random probability vectors. The vectors express the probability that an observation belongs to one of five classes.

numObs = 10; numClasses = 5; prob = rand(numObs,numClasses); tot = sum(prob,2); prob = prob./tot

`prob = `*10×5*
0.2938 0.0568 0.2365 0.2546 0.1582
0.3895 0.4174 0.0154 0.0137 0.1641
0.0427 0.3217 0.2854 0.0931 0.2573
0.2878 0.1529 0.2943 0.0145 0.2505
0.2640 0.3341 0.2834 0.0405 0.0780
0.0422 0.0614 0.3280 0.3564 0.2120
0.1078 0.1632 0.2876 0.2689 0.1725
0.1940 0.3249 0.1392 0.1125 0.2293
0.2356 0.1949 0.1613 0.2338 0.1745
0.3345 0.3326 0.0593 0.0119 0.2616

Define the set of five classes.

classes = ["Red","Yellow","Green","Blue","Purple"];

Decode the probabilities into the most probable classes by using the `onehotdecode`

function. The probability vectors are encoded into the second dimension (each column corresponds to a unique class), so specify the dimension containing encoded probabilities as `2`

. Obtain the most probable classes as a string vector.

`result = onehotdecode(prob,classes,2,"string")`

`result = `*10x1 string*
"Red"
"Yellow"
"Yellow"
"Green"
"Yellow"
"Blue"
"Green"
"Yellow"
"Red"
"Red"

### One-Hot Decode Classification Scores

Decode predicted class scores into predicted labels.

Load the `fisheriris`

data set. Create `X`

as a numeric matrix that contains four petal measurements for 150 irises. Create `S`

as a vector of categorical labels that contains the corresponding iris species.

```
load fisheriris
X = meas;
S = categorical(species);
```

One-hot encode the labels into probability vectors by using the `onehotencode`

function. Encode the probability vectors into the second dimension.

Y = onehotencode(S,2);

Compute the fitted coefficients of a simple linear classifier.

B = X\Y

`B = `*4×3*
0.0834 0.2117 -0.1481
0.2533 -0.3059 0.1412
-0.2270 0.1888 0.0181
-0.0635 -0.5749 0.5873

Predict the class scores from the fitted coefficients, and ensure that the scores are in the range [0,1].

scores = X*B; scores = min(1,max(0,scores));

Decode the predicted class scores into predicted labels by using the `onehotdecode`

function. Then, create a confusion chart to compare the true labels `S`

with the predicted labels `label`

.

label = onehotdecode(scores,categories(S),2); confusionchart(S,label)

## Input Arguments

`B`

— Probability vectors

numeric array

Probability vectors to decode, specified as a numeric array.

Values in `B`

must be between `0`

and
`1`

. If a probability vector in `B`

contains
`NaN`

values, the function decodes that observation to the class
with the largest probability that is not `NaN`

. If an observation
contains only `NaN`

values, the function decodes that observation to
the first class label in `classes`

.

**Data Types: **`single`

| `double`

`classes`

— Classes

cell array | string vector | numeric vector | character array

Classes, specified as a cell array of character vectors, a string vector, a numeric vector, or a two-dimensional character array.

**Data Types: **`single`

| `double`

| `int8`

| `int16`

| `int32`

| `int64`

| `uint8`

| `uint16`

| `uint32`

| `uint64`

| `string`

| `cell`

| `char`

`featureDim`

— Dimension containing probability vectors

positive integer

Dimension containing probability vectors, specified as a positive integer.

Use `featureDim`

to specify the dimension in `B`

that
contains the probability vectors. The function replaces each vector in
`B`

along the specified dimension with the element of
`classes`

in the same position as the highest value along the
vector.

The dimension of `B`

specified by `featureDim`

must have length equal to the number of classes specified by `classes`

.

`typename`

— Data type of decoded labels

`'categorical'`

(default) | character vector | string scalar

Data type of decoded labels, specified as a character vector or a string scalar.

Valid values of `typename`

are `'categorical'`

,
`'string'`

, and numeric types such as `'single'`

and `'int64'`

. If you specify a numeric type,
`classes`

must be a numeric vector.

**Example: **`'double'`

**Data Types: **`char`

| `string`

## Output Arguments

`A`

— Decoded class labels

categorical array (default) | string array | numeric array

Decoded class labels, returned as a categorical array, a string array, or a numeric array.

## Version History

**Introduced in R2021b**

## See Also

## Open Example

You have a modified version of this example. Do you want to open this example with your edits?

## MATLAB Command

You clicked a link that corresponds to this MATLAB command:

Run the command by entering it in the MATLAB Command Window. Web browsers do not support MATLAB commands.

# Select a Web Site

Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .

You can also select a web site from the following list:

## How to Get Best Site Performance

Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.

### Americas

- América Latina (Español)
- Canada (English)
- United States (English)

### Europe

- Belgium (English)
- Denmark (English)
- Deutschland (Deutsch)
- España (Español)
- Finland (English)
- France (Français)
- Ireland (English)
- Italia (Italiano)
- Luxembourg (English)

- Netherlands (English)
- Norway (English)
- Österreich (Deutsch)
- Portugal (English)
- Sweden (English)
- Switzerland
- United Kingdom (English)