Compare 8 distance metrices and 3 metric learning method in KNNClassificating an image dataset which has been extracted features by deep learning
Homework(project) from SJTU data science lesson CS
Sklearn:https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html#sklearn.neighbors.KNeighborsClassifier
Metric-learn:https://github.com/metric-learn/metric-learn
Metric-learn document:http://metric-learn.github.io/metric-learn/
https://cvml.ist.ac.at/AwA2/. This dataset consists of 37322 images of 50 animal classes with pre-extracted deep learning features for each image.
The name is borrowed from soft(hard)-margin SVM. soft KNN algorithm predicts one data by the distance of its nearest k data, while hard KNN predicts one by the mode label of its nearest k data.