Training

Model Deployment and Solo Scripting

Course Description

Many applications of chemometrics result in a model with the intent to use it in production or testing environments. Solo_Predictor and Model_Exporter are the two main solutions to deploying models built in PLS_Toolbox or Solo. This course describes these two very different approaches to model deployment and provides practical, hands-on examples of how to use each approach, and the situations where each is more appropriate. Applications of Solo_Predictor (and Solo) can be expanded through the use of Solo Scripting, a simple, flexible scripting language which can send instructions to load data, apply models and retrieve results.

Solo_Predictor is a stand-alone server which permits previously-built Eigenvector models to be used to make predictions on new data supplied by a user. We will describe Solo_Predictor installation and configuration to suit the commonest usage needs, and available client communication methods. Its abilities will be detailed, especially its versatile Solo Scripting language, its web-browser monitoring interface, and troubleshooting issues. Hands-on exercises will include communicating with a live Solo_Predictor server for different types of model application and extraction of relevant prediction features. Advanced Solo Scripting for complex scenarios will be introduced and demonstrated for use with Solo_Predictor and Solo (including variants).

Model_Exporter is an add-on feature to PLS_Toolbox or Solo which enables the export or translation of models into text form based on an interpretable format (MATLAB .m, Python .py, or XML) for use outside of these products. These exported models can be used in MATLAB or Python with minimal additional resources or programming language environments such as C# or Java using our provided interpreters for XML output.

The course includes hands-on computer time using MATLAB and PLS_Toolbox for participants to understand better the differences between the various options. A server will be set up in class so that participants can practice communicating with it and do live examples of online predictions.

Prerequisites

Experience building and saving models in PLS_Toolbox or Solo

Course Outline

  • Overview of model deployment approaches
    • Why Deploy Models?
    • Inherent Challenges
  • Solo_Predictor
    • Overview of Solo_Predictor as a prediction server
    • Installation and configuration
    • Usage methods (sockets, HTTP, wait for file, timer-based action)
    • Scripting language
      • Why script?
      • Typical script exchanges
      • Workspace objects
      • Script commands
      • Importing commands
      • Application of models and preprocessing
      • Requesting return values
      • Write to file
      • Other commands
    • Field Monitor
    • How to troubleshoot problems
  • Advanced Solo Scripting
  • Model_Exporter
    • Overview of Model_Exporter and supported methods
    • Export formats (MATLAB .m, Python .py, and XML)
    • Usage examples
      • MATLAB .m
      • Python .py
      • XML interpreters for Java or C#
  • Comparison of features and performance
    • Situations where Solo_Predictor is preferred
    • Situations where Model_Exporter is preferred
  • Live exercises
  • Conclusions