Amazon is ready to take the fight to IBM Watson Analytics' artificial intelligence through an easy-to-implement machine learning service geared toward developers.
Unveiled on Thursday, Amazon Machine Learning gives developers the chance to extract much more meaningful insights from their Amazon Web Services (AWS) hosted data without needing a background in data science and it is charged on a pay-as-you-go basis.
Developers are required to create a model, which can be done using Amazon Machine Learning's APIs and wizards combined with data stored on Amazon S3, Redshift or RDS. Developers can then use that to apply predictions and extract data, plus they can alter variables inside the wizards in order to draw the most appropriate predictions possible.
The system supports a trio of distinct predictions to cover all bases. Binary classification predicts one of two outcomes (e.g. a yes or no), multi-class classification can predict three or more outcomes (e.g. is this product a book, a movie, or an article of clothing?), and regression is able to predict a number (e.g. how many 27-inch monitors should we place in inventory?).
Following Microsoft's lead
Predictions are priced in two ways. Batch predictions, which covers a group of predictions provided at once, cost $0.10 per 1,000 predictions. Real-time predictions, which can handle predictions on demand, cost $0.10 per 1,000 but also incur a charge of $0.001 per hour for every 10MB provisioned for your model.
Amazon's decision to enter the machine learning is a clear nod to the success enjoyed by Microsoft's Azure Machine Learning, launched in July 2014, and IBM's renowned Watson Analytics service that has been growing quickly.
from http://bit.ly/1JwRBcI
