20.500.12556/RUL-160088 Investigating the interpretability of biometric face templates using gated sparse autoencoders and differentiable image parametrizations State-of-the-art face recognition models rely on deep, complex neural net architectures that produce relatively compact template vectors, making their mechanisms of operation difficult to interpret and understand. Recently, mechanistic interpretability has emerged as a promising approach to explain large language models. In this paper, we aim to apply such approaches to explain face recognition models. Our method involves transforming face image templates into sparse representations and analyzing their components by identifying images that maximize activation. Our results demonstrate that existing mechanistic interpretability techniques generalize well to previously unconsidered tasks and architectures, and that differentiable image parametrizations can serve as a useful additional means of confirming the interpretation of sparse representations. biometry face recognition sparse auto-encoders biometrija razpoznavanje obrazov redki samo-kodirniki true false true Angleški jezik Slovenski jezik Neznano 2024-08-19 09:07:15 2024-08-19 09:07:21 2024-08-20 03:46:12 0000-00-00 00:00:00 2024 0 0 5 str. 0000-00-00 Zaloznikova Objavljeno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.93'1 204387843 204384771 77_Investigating_the_Interpret.pdf 77_Investigating_the_Interpret.pdf 1 B0FAD6C235C3336E35B51A2E270450F9 b07c82eb4266ca871bc88d1b71e01f3c999e38c749dc568594e77b771f1b52c2 e34f691c-5df9-11ef-8f74-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=188540 https://openreview.net/forum?id=kUGkpykJdh 1 7cccc9aa-5df9-11ef-8f74-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=188539 Fakulteta za elektrotehniko 0 0 0