Linux/Unix>> MNIST-Classification>> 返回
项目作者： Byte7

项目描述 ：
MNIST classification using scikit-learn
高级语言： Jupyter Notebook
项目地址: git://github.com/Byte7/MNIST-Classification.git
创建时间： 2017-12-03T15:55:02Z
项目社区：https://github.com/Byte7/MNIST-Classification
开源协议：MIT License
下载


MNIST-Classification

MNIST classification using scikit-learn
Classifiers used are -

  1. SGDClassifier(for binary classification) - 98% accuracy
  2. RandomForestClassifier - 95% accuracy
  3. KNeighborsClassifier - 97% accuracy

Improvements

  1. Over 97% accuracy can be achieved using KNeighborsClassifier with good hyperparameters values(grid search on weights and n_neighbors hyperparameters)(takes long time)
  2. CNN can be used to get over 99%