Written by Adrian Bridgwater
A good proportion of the intelligence being engineered into the Internet of Things is classified into the technology sub-genre that we like to call machine learning.
Nothing too explicit thanks
If we understand machine learning to be the ability for devices to gain knowledge without being ‘explicitly programmed’, it is easy to see how useful this will be given the nature of IoT installations with deeply embedded units that may only enjoy occasional connection to the web for updates and data interchange.
Among the newer names in this space (not quite in the usual suspect corporate behemoth category) is Nimbix.
The firm is essentially a HPC cloud platform provider with a machine learning platform called Jarvice that offers ‘turnkey workflows’ i.e. essentially deployment templates aimed at reducing time to deployment from weeks to hours.
Built on Nvidia Graphic Processing Units (GPU’s), the machine learning prowess here is described as suitable for optimal neural network training.
Experienced machine learning developer, Hugh Perkins, author of the popular open source OpenCL libraries DeepCL and cltorch, is an avid user of the Nimbix cloud. Perkins says chose to work with Nimbix in addressing machine learning due to the powerful platform API, industry-leading selection of GPUs, superior-performance and economics.
Selling IoT neural brains
The Nimbix cloud platform is democratized and developer-friendly, allowing users to monetize their trained neural networks in the application marketplace. In other words, an application developer targeting the IoT space can start to build machine learning Artificial Intelligence that suits a specific deployment scenario and then take that intelligence to market as a separate defined commodity that can be resold to other use cases.