I have tried implementing a own classifier. I this classifier, I am taking the instances and creating an average template that I have used to compute the distance with the query instance. The label of the tempate with the closest distance is then given to the query instance.
MNIST dataset was used for this classifier.
The dataset can be found here .
The notebook contains the implementation of two algorithm from scratch. First is the Binary Relevance Classifier and second is the Classifier Chains.
Click here to have HTML preview.