eprintid: 10119535 rev_number: 14 eprint_status: archive userid: 608 dir: disk0/10/11/95/35 datestamp: 2021-01-27 14:24:14 lastmod: 2021-12-20 00:53:53 status_changed: 2021-01-27 14:24:14 type: proceedings_section metadata_visibility: show creators_name: Alzantot, M creators_name: Widdicombe, A creators_name: Julier, S creators_name: Srivastava, M title: NeuroMask: Explaining Predictions of Deep Neural Networks through Mask Learning ispublished: pub divisions: UCL divisions: B04 divisions: C05 divisions: F48 keywords: Computational modeling , Predictive models, Training, Cost function, Neural networks, Computer science, Image recognition note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. abstract: Deep Neural Networks (DNNs) deliver state-of-the-art performance in many image recognition and understanding applications. However, despite their outstanding performance, these models are black-boxes and it is hard to understand how they make their decisions. Over the past few years, researchers have studied the problem of providing explanations of why DNNs predicted their results. However, existing techniques are either obtrusive, requiring changes in model training, or suffer from low output quality. In this paper, we present a novel method, NeuroMask, for generating an interpretable explanation of classification model results. When applied to image classification models, NeuroMask identifies the image parts that are most important to classifier results by applying a mask that hides/reveals different parts of the image, before feeding it back into the model. The mask values are tuned by minimizing a properly designed cost function that preserves the classification result and encourages producing an interpretable mask. Experiments using state-of-art Convolutional Neural Networks for image recognition on different datasets (CIFAR-10 and ImageNet) show that NeuroMask successfully localizes the parts of the input image which are most relevant to the DNN decision. By showing a visual quality comparison between NeuroMask explanations and those of other methods, we find NeuroMask to be both accurate and interpretable. date: 2019-01-01 date_type: published publisher: IEEE official_url: https://doi.org/10.1109/SMARTCOMP.2019.00033 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1682597 doi: 10.1109/SMARTCOMP.2019.00033 lyricists_name: Julier, Simon lyricists_id: SJULI23 actors_name: Julier, Simon actors_id: SJULI23 actors_role: owner full_text_status: public publication: 2019 IEEE INTERNATIONAL CONFERENCE ON SMART COMPUTING (SMARTCOMP 2019) place_of_pub: Washington, DC, USA pagerange: 81-86 pages: 6 event_title: 5th IEEE International Conference on Smart Computing (SMARTCOMP) event_location: Washington, DC event_dates: 12 June 2019 - 14 June 2019 institution: 5th IEEE International Conference on Smart Computing (SMARTCOMP) book_title: 2019 IEEE International Conference on Smart Computing (SMARTCOMP) citation: Alzantot, M; Widdicombe, A; Julier, S; Srivastava, M; (2019) NeuroMask: Explaining Predictions of Deep Neural Networks through Mask Learning. In: 2019 IEEE International Conference on Smart Computing (SMARTCOMP). (pp. pp. 81-86). IEEE: Washington, DC, USA. Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10119535/1/1908.04389v1.pdf