Machine Learning on Paas
What is the importance of PaaS?
Why PaaS for ML
ML platform implementations typically require high configuration such as 128GB RAM and 2 TB hard disk(Standard). It would cost a lot to provide such resources on the servers. Moreover, if the servers are allocated to ML & AL, then the resources cannot be used for any other functionality. PaaS however can overcome this problem. While the resources can be made available to ML platforms during the time the algorithms are running, at other times they can be freely made available to other programs.
The other advantage with PaaS is its portability. Consider the scenario where a user has to make a prototype based on some requirement. When it is time to share the prototype with the clients, all the required packages, runtimes, libraries, dependencies etc will have to be installed on machine where the application needs to run.
PaaS on the other hand uses container technology, which is light weight. Once ready, the prototype can be committed to a docker image which has all the dependencies. Now the image can be freely shared with others without requiring to install any other software.
Advantages of PaaS
PaaS must mature its offerings in the following areas to harness the ML & AI capabilities even further.
1) Native Cloud service
It enables executing and managing traditional machine learning models ex: clustering, Regression and classification.
2) Vision analytics services
Need services that provide understanding of the images and content
3) Speech Processing services
Need services speech analytics capabilities such as translation and conversion to text
With the increasing use of Cloud Technology for data & its analysis, ML & AI on PaaS will help in bringing a great synergy between the two.