<< Click to Display Table of Contents >> Navigation:  Using SMILE Wrappers > Inference > Noisy-MAX decomposition

Noisy-MAX nodes can be handled more efficiently by enabling SMILE’s Noisy-MAX decomposition. When decomposition is active, the relevance reasoning layer automatically transforms Noisy-MAX nodes into a form that reduces the complexity of inference, often yielding substantial performance improvements. The behavior of the decomposition can also be fine-tuned by setting a limit on the number of parents allowed in the temporary structures created during inference. The following methods let you check whether decomposition is enabled, toggle it on or off, and configure the decomposition limit:

Python

is_noisy_decomposition_enabled() -> bool

set_noisy_decomposition_enabled(enable: bool) -> None

get_noisy_decomposition_limit() -> int

set_noisy_decomposition_limit(limit: int) -> None

Java

boolean isNoisyDecompositionEnabled();

void setNoisyDecompositionEnabled(boolean enable);

int getNoisyDecompositionLimit();

void setNoisyDecompositionLimit(int limit);

C#

bool IsNoisyDecompositionEnabled { get; set; }

int NoisyDecompositionLimit { get; set; }

R

enabled <- isNoisyDecompositionEnabled()

setNoisyDecompositionEnabled(enable)

limit <- getNoisyDecompositionLimit()

setNoisyDecompositionLimit(limit)