How to design and implement ML algorithms as opposed to model the mathematical properties of those algorithms.
Focus on issues concerning the design and implementation of
- Datasets
- Classes of ML algorithms that use those datasets
- Deployment of those algorithms
Engineering of ML includes
- Understanding requirements
- Collecting and curating datasets
- Pros and cons of each class of algorithms
- Deployment in production with a given environment and budget.
MLE includes any activity that lets machine learning algorithms be implemented as a part of an effective production system.
ML engeneers might rewrite and scale a data analyst’s code, packaging the code into an easy-to-deploy versioned package, optimizing the machine learning algorithm to make sure that it generates a model compatible with, and running correctly in, the organization’s production environment.
Data analysts may execute some of the MLE tasks, such as data collection, transformation, and feature engineering.
ML engineers often execute some of the data analysis tasks, including
- Selection of the learning algorithm
- Hyperparameter tuning
- Model evaluation.
#Goal
Given a customer with a set of requirements understand whether and how a specific ML algorithm will effectively and efficiently satisfy those requirements.
- Should we use ml? When to (not) Use ML
- Will ML deliver the expected value? ML Impact
- What engineering process will be needed?
- How much will it cost to maintain? ML Cost
Machine Learning