[160422]readme

Result:

a. perceptron training using spark on linearb (local mode 8 threads):

10 iteration: Total Training Time: 11554.2126369 Unlabeled accuracy: 0.901385180840 Unlabeled attachment accuracy: 0.905329456251

b. perceptron training using spark on the hadoop cluster(8 nodes):

5 iteration: Total Training Time: 7034.87671685 Unlabeled accuracy: 0.901205598190 Unlabeled attachment accuracy: 0.905157056329

10 iteration: Total Training Time: 15499.6485689 Unlabeled accuracy: 0.900900307685 Unlabeled attachment accuracy: 0.904863976462