{"id":{"repo_id":"rowan","oai_identifier":"oai:rdw.rowan.edu:etd-3804"},"canonical_url":"https://search.ndltd.org/etd/rowan/oai:rdw.rowan.edu:etd-3804","repository":{"repo_id":"rowan","name":"Rowan University","base_url":"https://rdw.rowan.edu/do/oai/"},"display":{"title":"Towards machine self-awareness - A Bayesian framework for uncertainty propagation in deep neural networks","abstract":"Deep neural networks (DNNs) have surpassed human-level accuracy in various fields, including object recognition and classification. However, DNNs being inherently deterministic, are unable to evaluate their confidence in the decisions. Bayesian inference provides a principled approach to reason about model confidence or uncertainty by estimating the posterior distribution of the unknown parameters. The challenge in DNNs is the multi-layer stages of non-linearities, which makes propagation of high-dimensional distributions mathematically intractable. This dissertation establishes the theoretical and algorithmic foundations of uncertainty or belief propagation by developing new deep learning models that can quantify their uncertainty in the decision and self-assess their performance. We introduce Tensor Normal distributions as priors over the network parameters and derive a first-order Taylor series mean-covariance propagation framework. We subsequently extend the first-order approximation to an unscented framework that propagates sigma points through the model layers and is shown to be accurate to at least the second-order approximation. The uncertainty in the output decision is given by the propagated covariance of the predictive distribution. Experiments on benchmark datasets demonstrate: 1) superior robustness against Gaussian noise and adversarial attacks; 2) self-assessment through predictive confidence that decreases quickly with increasing levels of ambient noise or attack; and 3) an ability to detect a targeted attack from ambient noise.","abstract_html":"Deep neural networks (DNNs) have surpassed human-level accuracy in various fields, including object recognition and classification. However, DNNs being inherently deterministic, are unable to evaluate their confidence in the decisions. Bayesian inference provides a principled approach to reason about model confidence or uncertainty by estimating the posterior distribution of the unknown parameters. The challenge in DNNs is the multi-layer stages of non-linearities, which makes propagation of high-dimensional distributions mathematically intractable. This dissertation establishes the theoretical and algorithmic foundations of uncertainty or belief propagation by developing new deep learning models that can quantify their uncertainty in the decision and self-assess their performance. We introduce Tensor Normal distributions as priors over the network parameters and derive a first-order Taylor series mean-covariance propagation framework. We subsequently extend the first-order approximation to an unscented framework that propagates sigma points through the model layers and is shown to be accurate to at least the second-order approximation. The uncertainty in the output decision is given by the propagated covariance of the predictive distribution. Experiments on benchmark datasets demonstrate: 1) superior robustness against Gaussian noise and adversarial attacks; 2) self-assessment through predictive confidence that decreases quickly with increasing levels of ambient noise or attack; and 3) an ability to detect a targeted attack from ambient noise.","abstract_has_math":false,"creators":["Dera, Dimah"],"institution":null,"degree_name":"PhD Doctor of Philosophy","degree_level":"Dissertation","degree_discipline":"Electrical and Computer Engineering","degree_department":null,"school":null,"contributors":["Bouaynaya, Nidhal","Rasool, Ghulam","Ramachandran, Ravi"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":null,"date_issued":"","date_published":null,"updated_at":"2026-11-19T08:59:12Z","subjects":["Convolutional Neural Networks","Deep neural networks","Tensor Normal Distribution","Unscented Transformation","Variational Density Propagation","Variational Inference","Neural networks (Computer science)","Uncertainty--Mathematical models","Electrical and Computer Engineering"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://rdw.rowan.edu/etd/2802","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://rdw.rowan.edu/etd/2802","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://rdw.rowan.edu/do/oai/?verb=GetRecord&metadataPrefix=oai_etdms&identifier=oai%3Ardw.rowan.edu%3Aetd-3804","prefix":"oai_etdms"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Bouaynaya, Nidhal","Rasool, Ghulam","Ramachandran, Ravi"]},{"key":"dc:creator","label":"Author","values":["Dera, Dimah"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2022-06-05T07:00:00Z"]},{"key":"dc:date.created","label":"Dc Date Created","values":["2020-06-04T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical and Computer Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["PhD Doctor of Philosophy"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Convolutional Neural Networks","Deep neural networks","Tensor Normal Distribution","Unscented Transformation","Variational Density Propagation","Variational Inference","Neural networks (Computer science)","Uncertainty--Mathematical models","Electrical and Computer Engineering"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://rdw.rowan.edu/etd/2802"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Deep neural networks (DNNs) have surpassed human-level accuracy in various fields, including object recognition and classification. However, DNNs being inherently deterministic, are unable to evaluate their confidence in the decisions. Bayesian inference provides a principled approach to reason about model confidence or uncertainty by estimating the posterior distribution of the unknown parameters. The challenge in DNNs is the multi-layer stages of non-linearities, which makes propagation of high-dimensional distributions mathematically intractable. This dissertation establishes the theoretical and algorithmic foundations of uncertainty or belief propagation by developing new deep learning models that can quantify their uncertainty in the decision and self-assess their performance. We introduce Tensor Normal distributions as priors over the network parameters and derive a first-order Taylor series mean-covariance propagation framework. We subsequently extend the first-order approximation to an unscented framework that propagates sigma points through the model layers and is shown to be accurate to at least the second-order approximation. The uncertainty in the output decision is given by the propagated covariance of the predictive distribution. Experiments on benchmark datasets demonstrate: 1) superior robustness against Gaussian noise and adversarial attacks; 2) self-assessment through predictive confidence that decreases quickly with increasing levels of ambient noise or attack; and 3) an ability to detect a targeted attack from ambient noise."]},{"key":"dc:title","label":"Title","values":["Towards machine self-awareness - A Bayesian framework for uncertainty propagation in deep neural networks"]}]}],"canonical_facts":{"dc:contributor":["Bouaynaya, Nidhal","Rasool, Ghulam","Ramachandran, Ravi"],"dc:creator":["Dera, Dimah"],"dc:date.available":["2022-06-05T07:00:00Z"],"dc:date.created":["2020-06-04T07:00:00Z"],"dc:description.abstract":["Deep neural networks (DNNs) have surpassed human-level accuracy in various fields, including object recognition and classification. However, DNNs being inherently deterministic, are unable to evaluate their confidence in the decisions. Bayesian inference provides a principled approach to reason about model confidence or uncertainty by estimating the posterior distribution of the unknown parameters. The challenge in DNNs is the multi-layer stages of non-linearities, which makes propagation of high-dimensional distributions mathematically intractable. This dissertation establishes the theoretical and algorithmic foundations of uncertainty or belief propagation by developing new deep learning models that can quantify their uncertainty in the decision and self-assess their performance. We introduce Tensor Normal distributions as priors over the network parameters and derive a first-order Taylor series mean-covariance propagation framework. We subsequently extend the first-order approximation to an unscented framework that propagates sigma points through the model layers and is shown to be accurate to at least the second-order approximation. The uncertainty in the output decision is given by the propagated covariance of the predictive distribution. Experiments on benchmark datasets demonstrate: 1) superior robustness against Gaussian noise and adversarial attacks; 2) self-assessment through predictive confidence that decreases quickly with increasing levels of ambient noise or attack; and 3) an ability to detect a targeted attack from ambient noise."],"dc:identifier":["https://rdw.rowan.edu/etd/2802"],"dc:subject":["Convolutional Neural Networks","Deep neural networks","Tensor Normal Distribution","Unscented Transformation","Variational Density Propagation","Variational Inference","Neural networks (Computer science)","Uncertainty--Mathematical models","Electrical and Computer Engineering"],"dc:title":["Towards machine self-awareness - A Bayesian framework for uncertainty propagation in deep neural networks"],"thesis:degree_discipline":["Electrical and Computer Engineering"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["PhD Doctor of Philosophy"]},"updated_at":"2026-11-19T08:59:12Z"}