From 2396bcdd6cec9324d31547c192ec3d13bfd08d05 Mon Sep 17 00:00:00 2001 From: annahedstroem Date: Fri, 24 Nov 2023 12:11:56 +0100 Subject: [PATCH] add back sys typo --- quantus/metrics/randomisation/smooth_mprt.py | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/quantus/metrics/randomisation/smooth_mprt.py b/quantus/metrics/randomisation/smooth_mprt.py index db84cab3..4885b4b7 100644 --- a/quantus/metrics/randomisation/smooth_mprt.py +++ b/quantus/metrics/randomisation/smooth_mprt.py @@ -6,6 +6,7 @@ # You should have received a copy of the GNU Lesser General Public License along with Quantus. If not, see . # Quantus project URL: . +import sys import warnings from typing import ( Any, @@ -52,9 +53,9 @@ class SmoothMPRT(Metric): """ Implementation of the Smooth MPRT by Hedström et al., 2023. - The Sampling Model Parameter Randomisation measures the distance between the original attribution and a newly computed - attribution throughout the process of cascadingly/independently randomizing the model parameters of one layer - at a time. + The Smooth Model Parameter Randomisation adds a "denoising" preprocessing step to the original MPRT, + where the explanations are averaged over N noisy samples before the similarity between the original- + and fully random model's explanations is measured. References: 1) Hedström, Anna, et al. "Sanity Checks Revisited: An Exploration to Repair the Model Parameter