With unprecedented growth in sensors and Internet of Things, the world we live has enormous amount of data that are being captured. The data either in the form of image, discrete data or audio is being transmitted to server for analysis (Mostly OLAP into Hadoop like platform). The IOTs are being treated as just as a data capture and analytics is outsourced to server farm. Does IOT needs to be just a data capture, can pattern creation and deep learning done at IOT level
To frame this question in different form, can the processing be done “In-memory”. Can system look for new patterns with very small footprint.
H2O.ai provides the solution; this stack is the best ML that can run on JVM. With small footprint, each IOT can now do more than just data capture.
H2O has 3 major advantages for IOT development in Deep Learning.
- Develop new Pattern in memory and compare for anomoly
- Small footprint and fast deployment
- Basic level security at the data capture level.
Having these capabilities opens new door for IOT manufacturer and service provider. They can respond fast to scenarios faster and make decision at real time level.
Provide more value add to the consumer and emerge with new patterns.
My reason to go for H2O are
- Open source
- Integration with existing development environment like R and Python
- Running on small footprint
- Good community support.
Check out H2O at H2O.ai
The future is to have more learning capabilities at the data capture level. We expect more real time detection of anomalies from the IOT sensors.
