TransJury: Towards Explainable Transfer Learning through Selection of Layers from Deep Neural Networks

Arefeen, Md Adnan; Nimi, Sumaiya Tabassum; Uddin, Yusuf Sarwar; Lee, Yugyung

doi:10.1109/bigdata52589.2021.9671723

Cited by 5 publications

(4 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a universal feature is shared for all downstream tasks, it is computationally efficient but raises a privacy concern while sharing with outside agents due to offering a common feature for all tasks. Similar behavior patterns can be found in other recent literature [2,3] where features from multiple layers of deep models are fused to form the universal features and image classification task is accomplished.…”

Section: Motivation and Challengessupporting

confidence: 73%

“…Without loss of generality, we use the same classifier model size for all tasks. After training of encoder with DP and the private classifier, we use the task-privacy loss to train the classifier for smile for case (1) and the smile and cheekbone classifier jointly for case (2). Then, we test the performance of all tasks and task privacy by interchanging the task-metamorphosis modules.…”

Section: Resultsmentioning

confidence: 99%

“…The metamorphosis module is shown in Figure 3 with the shape for each layer. To make the task features distinct, we use a similarity metric as used in recent literature [2,18]. In this case, we use the SSIM metric to compare the structural similarity among task features, and in the loss function, it learns to project them far from each other based on the weight given to this metric.…”

Section: Task Privacymentioning

confidence: 99%

See 2 more Smart Citations

MetaMorphosis: Task-oriented Privacy Cognizant Feature Generation for Multi-task Learning

Arefeen

Uddin

et al. 2023

Proceedings of the 8th ACM/IEEE Conference on Internet of Things Design and Implementation

Self Cite

View full text Add to dashboard Cite

With the growth of computer vision applications, deep learning, and edge computing contribute to ensuring practical collaborative intelligence (CI) by distributing the workload among edge devices and the cloud. However, running separate single-task models on edge devices is inefficient regarding the required computational resource and time. In this context, multitask learning allows leveraging a single deep learning model for performing multiple tasks, such as semantic segmentation and depth estimation on incoming video frames. This single processing pipeline generates common deep features that are shared among multi-task modules. However, in a collaborative intelligence scenario, generating common deep features has two major issues. First, the deep features may inadvertently contain input information exposed to the downstream modules (violating input privacy). Second, the generated universal features expose a piece of collective information than what is intended for a certain task, in which features for one task can be utilized to perform another task (violating task privacy). This paper proposes a novel deep learning-based privacy-cognizant feature generation process called "MetaMorphosis" that limits inference capability to specific tasks at hand. To achieve this, we propose a channel squeeze-excitation based feature metamorphosis module, Cross-SEC, to achieve distinct attention of all tasks and a de-correlation loss function with differential-privacy to train a deep learning model that produces distinct privacy-aware features as an output for the respective tasks. With extensive experimentation on four datasets consisting of diverse images related to scene understanding and facial attributes, we show that MetaMorphosis outperforms recent adversarial learning and universal feature generation methods by guaranteeing privacy requirements in an efficient way for image and video analytics. CCS Concepts: • Computing methodologies → Computer vision tasks; Scene understanding; • Security and privacy → Privacy protections.

show abstract

Section: Motivation and Challengessupporting

confidence: 73%

Section: Resultsmentioning

confidence: 99%

Section: Task Privacymentioning

confidence: 99%

See 1 more Smart Citation

MetaMorphosis: Task-oriented Privacy Cognizant Feature Generation for Multi-task Learning

Arefeen

Uddin

et al. 2023

Proceedings of the 8th ACM/IEEE Conference on Internet of Things Design and Implementation

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although DTL-based ASR models have achieved great success, they are still perceived as "black box" schemes that lack interpretation. This does not provide convincing insights into "how" and "why" they can reach final decisions [191]. This can doubt the credibility of reached decisions and lack compelling evidence for convincing users or companies that these algorithms can work repeatedly.…”

Section: Interpretation Of Dtl Modelsmentioning

confidence: 96%

Video surveillance using deep transfer learning and deep domain adaptation: Towards better generalization

Himeur

Al‐Maadeed

Kheddar

et al. 2023

Engineering Applications of Artificial Intelligence

View full text Add to dashboard Cite

Deep transfer learning for automatic speech recognition: Towards better generalization

Kheddar

Himeur

Al-Maadeed

et al. 2023

Knowledge-Based Systems

View full text Add to dashboard Cite

TransJury: Towards Explainable Transfer Learning through Selection of Layers from Deep Neural Networks

Cited by 5 publications

References 23 publications

MetaMorphosis: Task-oriented Privacy Cognizant Feature Generation for Multi-task Learning

MetaMorphosis: Task-oriented Privacy Cognizant Feature Generation for Multi-task Learning

Video surveillance using deep transfer learning and deep domain adaptation: Towards better generalization

Deep transfer learning for automatic speech recognition: Towards better generalization

Contact Info

Product

Resources

About