Authoring a Hand-Written Digit Recognizer in Microsoft Azure Machine Learning Studio
In this lab, use Microsoft Azure Machine Learning to author an experiment for recognizing hand-written digits through a neural network multiclass classification algorithm. We use a labeled dataset with 5000 training examples of the handwritten digits to train and test a multiclass neural network with training data. Optionally, we also test the trained neural network model with cross validation data, evaluate the training and cross validate the models. We then save the trained model, create a scoring experiment and publish a scoring web service to a staging environment which will be used to predict the digit that corresponds to input data provided as parameter to the service.