A binary random variable is restricted to two categories, conventionally represented as 0 and 1. These values can encode states such as no/yes, not firing/firing, or do not sell/sell. section 7.2.1 section 7.2
Binary variables are a special case of categorical variables: each value identifies a category rather than a continuously varying quantity. In machine-learning systems, separate machines can emit binary decisions that another function combines. section 7.2.1 section 7.2
Common logical interpretations include:
- Logical OR produces a positive decision when at least one binary input is 1. section 7.2.1
- Logical AND produces a positive decision when the required inputs are all 1. section 7.2.1
- Logical NOT reverses a binary decision, changing 1 to 0 and 0 to 1. section 7.2.1
- A classifier can map continuous or categorical inputs to a binary output such as face or not face. section 7.2.2
Face detection
A face detector receives an image and outputs 1 when it identifies a face or 0 when it does not. The image may be represented as a high-dimensional input vector, while the classifier's output remains binary. section 7.2.2
Threat detection
Represent hardness and sharpness as binary inputs: 1 means the property is present and 0 means it is absent. A logical OR model marks an object threatening when either property is present; an AND model requires both. section 7.2.1
