I am a beginner at learning deep learning and recently I have got some work related to machine learning and deep learning but when I was learning ANN multi-class output method to predict the output. Y_predshowing an array with [[ True False False
False True True]] something like this. I want these outputs in just number for example as 0,1,2 then How is it possible to do? Which library is helpful to convert outputs in numbers? or is it not possible?
The term you're looking for is "One-hot encoding". If you have just a few classes you can directly say that an output [0 1 0] should be interpreted as class 2, or output [1 0 0] represents class 3. The other way is to use a library, such as sckit-learn or a built-in function from tensorlow or keras (In Pytorch you don't need to use one-hot encoding). Here's a reference with examples: https://towardsdatascience.com/categorical-encoding-using-label-encoding-and-one-hot-encoder-911ef77fb5bd
Dear Riken Mangukiya ,
Look the link, maybe useful.
https://towardsdatascience.com/first-neural-network-for-beginners-explained-with-code-4cfd37e06eaf
Regards,
Shafagat
The below link can be a good reference to develop such multiclass classification problems:
https://towardsdatascience.com/predictive-modeling-and-multiclass-classification-a4d2c428a2eb
You can encode this programmatically if you coding the NN yourself. Alternatively via settings commands if you are using predictive tools. Check the following links:
https://stackabuse.com/creating-a-neural-network-from-scratch-in-python-multi-class-classification/
https://visualstudiomagazine.com/articles/2020/12/15/pytorch-network.aspx
Javad Sharifi, actually your links are used to handling normalization of input data. What Riken Mangukiya is asking for is about one-hot encoding of the outputs, which is completely different. I just wanted to stress that out because that could create some future confusion.
Javad Sharifi , yes, but the question is simpler than that. The problem is not how to pre-process or handle input categorical data, the question is basically how to transform the binary OUTPUT of a multi-class classifier (e.g. [False, True, False, False]) into a numeric class (e.g. "2"). I think it's important to clarify that because, otherwise, some readers could think that some of the normalization techniques you mentioned have to be applied to the outputs as well.
use approach named 'one hot encoding' https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html
or 'get_dummies' https://pandas.pydata.org/docs/reference/api/pandas.get_dummies.html
True False False False True True = 1 0 0 0 1 1
It is binary written down 35
1*(2^5)+0*(2^4)+0*(2^3)+0*(2^2)+1*(2^1)+1*(2^0)=35
:-)
Doesn't it work?
Then you can cluster groups of numbers.
- Badges
- Science topic