How deeply does Machine Learning System Design cover soft target?

Defines mixed class membership and specific weights, while deferring a detailed implementation recipe.

Soft target

As explained in Machine Learning System Design

A label representation that assigns weighted or probabilistic membership rather than a single categorical class.

A soft target assigns pixels weighted membership in multiple classes instead of forcing a single categorical label. In fine-grained segmentation, a boundary pixel may belong partly to foreground and partly to background. section 3.1

Hair against a background

Image matting treats hair segmentation as a soft-label problem: difficult edge pixels receive specific foreground and background weights. The same idea was adapted to manufacturing data and reduced test error. section 3.1

How Machine Learning System Design covers it

liveBook manning.com

Read Machine Learning System Design on liveBook Get the book at manning.com

Related concepts

labelcontrasts with · A label is an assigned target value, category, or reference annotation associated with an item, used to identify its expected outcome or support learning and evaluation.
distillationused by · Knowledge distillation is a training technique in which a student model learns to reproduce or approximate the outputs or behavior of a teacher model.
teacher modelproduced by · A teacher model is a model whose outputs, labels, or learned behavior provide training signals or targets for another model.
student modeluses · A student model is a model trained to approximate or reproduce the outputs, capabilities, or behavior of another model that provides training guidance.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.