Merge pull request #1747 from hpe-dev-incubator/cms/blog/smart-use-cases-for-smartsim

Update Blog “smart-use-cases-for-smartsim”

Author: choukden

content/blog/smart-use-cases-for-smartsim.md

@@ -18,9 +18,9 @@ With SmartSim, you get an infrastructure library component and a client library
 
 The client library that's a part of SmartSim enables users to embed this machine learning capability inside of traditional applications.  The library is available in four languages, Python, C, C++ and Fortran. It provides a very simple API that allows the transfer of data between the traditional applications and the database. The simplicity of the API allows users to rapidly integrate their application with only a minimal amount of modification to the original source code (usually less than 10 lines).  
 
-One of the early use cases that the designers focused on was using SmartSim to enable better predictions inside of a global ocean model. The [MOM6 (Modular Ocean Model)](https://www.gfdl.noaa.gov/mom-ocean-model/) global ocean model has an algorithm that calculates a term called Eddy Kinetic Energy (EKE) which governs the strength of turbulence. It turns out that the parameterization is not very accurate. So, using SmartSim, we trained a machine learning model on a high fidelity run to learn the mapping from a coarsened version of the quantities (similar to those in the low fidelity run) to the true EKE. Not only did that machine learning model improve the accuracy by at least 20%, but there was not visible degradation in performance, even though this simulation required about 1.6 million inferences per second. 
+One of the early use cases that the designers focused on was using SmartSim to enable better predictions inside of a global ocean model. The [MOM6 (Modular Ocean Model)](https://www.gfdl.noaa.gov/mom-ocean-model/) global ocean model has an algorithm that calculates a term called Eddy Kinetic Energy (EKE) which governs the strength of turbulence. It turns out that the parameterization is not very accurate. So, using SmartSim, we trained a machine learning model on a high fidelity run to learn the mapping from a coarsened version of the quantities (similar to those in the low fidelity run) to the true EKE. Not only did that machine learning model improve the accuracy by at least 20%, but there was no visible degradation in performance, even though this simulation required about 1.6 million inferences per second. 
 
-SmartSim is great for users who have very specialized domain expertise in things like computational fluid dynamics, molecular dynamics but don’t necessarily have the knowledge set to deploy the machine learning infrastructure for high performance. SmartSim removes that technical barrier for users so that they can spend more time experimenting with combining scientific simulations and ML.
+SmartSim is great for users who have very specialized domain expertise in things like computational fluid dynamics or molecular dynamics; but don’t necessarily have the knowledge set to deploy the machine learning infrastructure for high performance. SmartSim removes that technical barrier for users so that they can spend more time experimenting with combining scientific simulations and ML.
 
 For more information on SmartSim, please reference the [SmartSim page](https://developer.hpe.com/platform/smartsim/home) found on the [HPE Developer Community portal](https://developer.hpe.com/). You can also access [SmartSim’s GitHub page](https://github.com/CrayLabs/SmartSim).