Machine Learning as a Service — Google Prediction API vs Amazon ML
Almost all big Giants are budding up for Machine learning as next big thing. Claiming the futuristic approach in technology and development, these companies have put their larger amount of effort and money in Big Data and Analytics through machine learning.
I started getting my hands around two giants Google with their Prediction API and Amazon with it’s AWS ML platform will be good enough to know what everyone else is or should be up to in ML solution building. So here is one by one feature comparison of both the brilliant solutions:
Cloud Based Solution
Amazon delivers it through its machine learning console. For data storage it uses S3 buckets. However there are other connectors as well which work good.
Google Prediction API can be accessed through an api exposed which needs authentication through P12 or Json files. Programmatically you can access the API while it has its
Prediction Methods
Google is exposed via apis and API explorer. Apart from that you depend on programming language to interact with it in order to train and predict. API explorer is more like a testing tool than a full blown operational console.
On the other hand Amazon has a robust console to create datasource, modelling, train and predict. From a single screen it provides you the mechanism to access all aspects of its machine learning solution.
Learning Curve
Google prediction is straightforward and if you are aware of data/model creation, you just need to download your favorite sdk in any language and get started with examples given in it.
Amazon ML needs you to understand everything from Amazon’s point of view. Starting from Model creation to predict. It takes longer to learn but invested time is worth it at the end.
Language Support
They both support Java, PHP, Go, .NET. While amazon supports C++ development preview, Google supports python as well.
Documentation
Amazon has rich documentation as it has lots of things to learn. Google has specific examples of client libraries according to your needs.
Stackoverflow has more questions on Google prediction api than Amazon’s ML.
Pricing
Google prediction API is free for first six months with restricted use of api calls. However that should be ok until you go live.
Amazon has pay per use plan which is billed start from the word go.
Conclusion
Amazon ML is good and proven old technology and has been there for a while in the market. Google is experimenting a lot with its solutions. Keeping the history in mine, we can expect Google to launch variety around its ML solutions as it is just the beginning, we know its Cloud ML platform is coming to beta this fall with full integration with a Vision, Speech, Cloud Storage and other services
Talking about data, Google is backed by huge data it possesses and hence we can expect many solutions in near future which will be benefitted by the ML solution it is going to evolve with.
My pick for tight deadline and versatile prediction results will be Amazon. However if I am more inclined towards creativity and innovation, Google is what I am looking for.
Originally published at Next Gen Technology Insider.