Historically, one of the perceived disadvantages of using R for data science projects was the difficulty in deploying machine learning models built in R. This often meant that companies used R mainly as a prototyping tool to build models which were then rewritten in another language, such as Java and .NET. It is also one of the main reasons cited for companies switching to Python for data science as Python has more glue code, which allows it to interface with other programming languages.
Thankfully, this is changing. One interesting new product from RStudio, called RStudio Connect, allows companies to create a platform for sharing R-Shiny applications, reports in R Markdown, dashboards, and models. This allows companies to serve machine learning models using a REST interface.
The TensorFlow (and Keras) models we have created in this book can be deployed...
