The field of artificial intelligence is moving towards increased transparency and accountability, with a focus on explainable AI (XAI) and efficient computing. Recent developments have led to the creation of novel frameworks and techniques that enable real-time outcome interpretations, energy-efficient hardware acceleration, and improved model interpretability. These advancements have significant implications for various applications, including edge devices, medical diagnosis, and safety-critical systems. Notably, the integration of XAI with approximate computing techniques has shown promising results, achieving improved energy efficiency while maintaining comparable accuracy. Noteworthy papers include ApproXAI, which proposes a framework for energy-efficient XAI using approximate computing, and EPSILON, which introduces a lightweight framework for adaptive fault mitigation in approximate deep neural networks. Additionally, the Dynamic Contextual Attention Network (DCAN) and eNCApsulate demonstrate innovative approaches to transforming spatial representations into adaptive insights for endoscopic polyp diagnosis and precision diagnosis on capsule endoscopes, respectively.
Explainable AI and Efficient Computing in Emerging Applications
Sources
ApproXAI: Energy-Efficient Hardware Acceleration of Explainable AI using Approximate Computing
Integrating Explainable AI for Energy Efficient Open Radio Access Networks
Explainable AI for UAV Mobility Management: A Deep Q-Network Approach for Handover Minimization
Periodic Online Testing for Sparse Systolic Tensor Arrays
Myocardial Region-guided Feature Aggregation Net for Automatic Coronary artery Segmentation and Stenosis Assessment using Coronary Computed Tomography Angiography
Systematic Hardware Integration Testing for Smart Video-based Medical Device Prototypes
EPSILON: Adaptive Fault Mitigation in Approximate Deep Neural Network using Statistical Signatures
Dynamic Contextual Attention Network: Transforming Spatial Representations into Adaptive Insights for Endoscopic Polyp Diagnosis
eNCApsulate: NCA for Precision Diagnosis on Capsule Endoscopes
Anomaly-Driven Approach for Enhanced Prostate Cancer Segmentation
AI-Assisted Decision-Making for Clinical Assessment of Auto-Segmented Contour Quality
Surrogate modeling of Cellular-Potts Agent-Based Models as a segmentation task using the U-Net neural network architecture