pulumi-google-native-kotlin/com.pulumi.googlenative.ml.v1.kotlin.inputs/GoogleCloudMlV1__ExplanationConfigArgs

GoogleCloudMlV1__ExplanationConfigArgs

data class GoogleCloudMlV1__ExplanationConfigArgs(val integratedGradientsAttribution: Output<GoogleCloudMlV1__IntegratedGradientsAttributionArgs>? = null, val sampledShapleyAttribution: Output<GoogleCloudMlV1__SampledShapleyAttributionArgs>? = null, val xraiAttribution: Output<GoogleCloudMlV1__XraiAttributionArgs>? = null) : ConvertibleToJava<GoogleCloudMlV1__ExplanationConfigArgs>

Message holding configuration options for explaining model predictions. There are three feature attribution methods supported for TensorFlow models: integrated gradients, sampled Shapley, and XRAI. /ai-platform/prediction/docs/ai-explanations/overview

Constructors

Link copied to clipboard
fun GoogleCloudMlV1__ExplanationConfigArgs(integratedGradientsAttribution: Output<GoogleCloudMlV1__IntegratedGradientsAttributionArgs>? = null, sampledShapleyAttribution: Output<GoogleCloudMlV1__SampledShapleyAttributionArgs>? = null, xraiAttribution: Output<GoogleCloudMlV1__XraiAttributionArgs>? = null)

Functions

Link copied to clipboard
open override fun toJava(): GoogleCloudMlV1__ExplanationConfigArgs

Properties

Link copied to clipboard
val integratedGradientsAttribution: Output<GoogleCloudMlV1__IntegratedGradientsAttributionArgs>? = null

Attributes credit by computing the Aumann-Shapley value taking advantage of the model's fully differentiable structure. Refer to this paper for more details: https://arxiv.org/abs/1703.01365

Link copied to clipboard
val sampledShapleyAttribution: Output<GoogleCloudMlV1__SampledShapleyAttributionArgs>? = null

An attribution method that approximates Shapley values for features that contribute to the label being predicted. A sampling strategy is used to approximate the value rather than considering all subsets of features.

Link copied to clipboard
val xraiAttribution: Output<GoogleCloudMlV1__XraiAttributionArgs>? = null

Attributes credit by computing the XRAI taking advantage of the model's fully differentiable structure. Refer to this paper for more details: https://arxiv.org/abs/1906.02825 Currently only implemented for models with natural image inputs.