Linear Learner Algorithm is a Supervised Learning algorithm that can be used to solve three types of problems: Binary classification; Multi-class classification; and Regression. The algorithm is trained with lists of data comprising a high dimensional vector x and a label y to learn the equation of the line.
The Linear Learner Algorithm uses Stochastic Gradient Descent to fit a line to the data points.
- AWS docs: https://docs.aws.amazon.com/sagemaker/latest/dg/linear-learner.html
|Data types and format||Tabular|
|Learning paradigm or domain||Supervised Learning|
|Problem type||Binary/multi-class classification, Regression|
|Use case examples||Predict a numeric/continuous value; Predict if an item belongs to a category|
|High dimensional vector||Numeric label|
|Binary classification||0 or 1|
|Multi-class classification||0 to num_class – 1|
- recordIO-wrapped protobuf
Model artifacts and inference
|Learning paradigm||Supervised Learning|
|Request format||recordIO-wrapped protobuf, CSV, JSON|
|Result||Regression: predicted score|
Binary Classification: predicted label 0 or 1 and the strength of the prediction.
Multi-class Classification: predicted label and a list of scores for each class showing the strength of the prediction.
For training the Linear Learner Algorithm single or multi-machine CPU and GPU instances can be used. The same can be used for inferencing with GPU providing little or no improvement over CPU instances.
Amazon SageMaker’s Built-in Algorithm Webinar Series: Linear Learner
This is a 58.54 minutes video from AWS.