Resilient Machine Learning Models and Optical Network Innovations

The current developments in the research area are primarily focused on enhancing the robustness and efficiency of machine learning models, particularly in the context of adversarial attacks and noisy data environments. There is a notable trend towards developing more resilient systems against hardware vulnerabilities and adversarial perturbations, with a specific emphasis on multi-task learning frameworks and optical neural network accelerators. Innovations in wavelength arbitration for microring-based transceivers and open research frameworks for optical data center networks are also advancing the field by addressing scalability and interoperability challenges. Additionally, methods for improving the resistance to noisy label fitting and proactive gradient conflict mitigation in multi-task learning are being explored, offering potential improvements in model generalization and performance. Notably, the integration of sparse training techniques with existing gradient manipulation methods shows promise in reducing gradient conflicts and enhancing overall model efficiency.

Sources

Scalable Wavelength Arbitration for Microring-based DWDM Transceivers

SafeLight: Enhancing Security in Optical Convolutional Neural Network Accelerators

Improving Resistance to Noisy Label Fitting by Reweighting Gradient in SAM

Adversarial Bounding Boxes Generation (ABBG) Attack against Visual Object Trackers

Stealthy Multi-Task Adversarial Attacks

Lighthouse: An Open Research Framework for Optical Data Center Networks

Proactive Gradient Conflict Mitigation in Multi-Task Learning: A Sparse Training Perspective

Built with on top of