[−][src]Crate sparrow
Sparrow is an implementation of TMSN for boosting.
From a high level, Sparrow consists of three components,
- scanner: it runs the boosting process, which scans the samples in memory, and updates the current model by finding a new weak rule to be added to the score function;
- sampler: it samples examples from disk and updates their scores according to the latest score function,
- model manager: it assigns tasks to the scanners, receives model updates from them, and maintains the current score function.
Modules
head | Implementation of the components running on head node, specifically the scanner and the model manager |
scanner | Implementation of the scanner |
Functions
testing | Test a model |
training | Train a model |