Classify the MNIST data by LIBSVM in Python.
Copyright (C) 2017 Jing Wang

The classification accuracy (93.99%) is terribly low compared to deep 
learning methods. However, it is a start. 
https://www.tensorflow.org/get_started/mnist/beginners

Data and softwares:
The MNIST data (http://yann.lecun.com/exdb/mnist/)
Python 3.5.2, https://www.continuum.io/downloads
TensorFlow 1.0.1, https://www.tensorflow.org/
Numpy 1.11.1
LIBSVM 3.22, https://github.com/cjlin1/libsvm
LIBSVM 3.21, http://www.lfd.uci.edu/~gohlke/pythonlibs/#libsvm (optional)

Usage:
1) Put the Python script under the libsvm-master/python directory
  if you use LIBSVM 3.22, https://github.com/cjlin1/libsvm.
2) Install LIBSVM 3.21, http://www.lfd.uci.edu/~gohlke/pythonlibs/#libsvm 
  (optional). By this way, you could run "from svmutil import *" anywhere.
Then run classify_MNIST.py.

2017-3-15 09:36:34