singlepp::ClassifySingleOptions< Float_ > Struct Template Reference

Options for classify_single() and friends. More...

#include <classify_single.hpp>

Public Attributes
Float_	quantile = 0.8
Float_	fine_tune_threshold = 0.05
bool	fine_tune = true
int	num_threads = 1

Detailed Description

template<typename Float_ = DefaultFloat>
struct singlepp::ClassifySingleOptions< Float_ >

Options for classify_single() and friends.

Template Parameters

Float_

Floating-point type for the correlations and scores.

Member Data Documentation

fine_tune

template<typename Float_ = DefaultFloat>

bool singlepp::ClassifySingleOptions< Float_ >::fine_tune = true

Whether to perform fine-tuning. This can be disabled to improve speed at the cost of accuracy.

fine_tune_threshold

template<typename Float_ = DefaultFloat>

Float_ singlepp::ClassifySingleOptions< Float_ >::fine_tune_threshold = 0.05

Score threshold to use to select the top-scoring subset of labels during fine-tuning. Larger values increase the chance of recovering the correct label at the cost of computational time.

This threshold should not be set to to a value that is too large. Otherwise, the first fine-tuning iteration would just contain all labels and there would be no reduction of the marker space.

num_threads

template<typename Float_ = DefaultFloat>

int singlepp::ClassifySingleOptions< Float_ >::num_threads = 1

Number of threads to use. The parallelization scheme is determined by tatami::parallelize().

quantile

template<typename Float_ = DefaultFloat>

Float_ singlepp::ClassifySingleOptions< Float_ >::quantile = 0.8

Quantile probability in [0, 1]. This is used to define a per-label score for each test cell, by applying it to the distribution of correlations between the test cell and that label's reference profiles. A value closer to 0.5 focuses on the behavior of the majority of a label's reference profiles. A smaller value will be more sensitive to the presence of a subset of profiles that are more similar to the test cell, which can be useful when the reference profiles themselves are heterogeneous.

---

The documentation for this struct was generated from the following file:

• singlepp/classify_single.hpp