MétaCan
Menu
Back to cohort

Building Trust in Deep Learning Models via a Self- Interpretable Visual Architecture

2023· article· en· W4388894069 on OpenAlex
Weimin Zhao, Qusay H. Mahmoud, Sanaa Alwidian

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

Venuenot available
Typearticle
Languageen
FieldComputer Science
TopicAdversarial Robustness in Machine Learning
Canadian institutionsOntario Tech University
Fundersnot available
KeywordsInterpretabilityComputer scienceArtificial intelligenceDeep learningRobustness (evolution)Machine learningConvolutional neural networkConsistency (knowledge bases)ArchitectureFeature (linguistics)Feature learningContext (archaeology)

Abstract

fetched live from OpenAlex

Deep learning models are being utilized and further developed in many application domains, but challenges still exist regarding their interpretability and consistency. Interpretability is important to provide users with transparent information that enhances the trust between the user and the learning model. It also gives developers feedback to improve the consistency of their deep learning models. In this paper, we present a novel architectural design to embed interpretation into the architecture of the deep learning model. We apply dynamic pixel-wised weights to input images and produce a highly correlated feature map for classification. This feature map is useful for providing interpretation and transparent information about the decision-making of the deep learning model while keeping full context about the relevant feature information compared to previous interpretation algorithms. The proposed model achieved 92% accuracy for CIFAR 10 classifications without finetuning the hyperparameters. Furthermore, it achieved a 20% accuracy under 8/255 PGD adversarial attack for 100 iterations without any defense method, indicating extra natural robustness compared to other Convolutional Neural Network (CNN) models. The results demonstrate the feasibility of the proposed architecture.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: Simulation or modeling
GenreCandidate signal: Methods · Consensus signal: none
Teacher disagreement score0.701
Threshold uncertainty score0.875

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0010.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0010.002
Science and technology studies0.0000.000
Scholarly communication0.0000.001
Open science0.0010.001
Research integrity0.0000.001
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.011
GPT teacher head0.269
Teacher spread0.258 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it