Researchers have introduced Deep4ge, a new benchmark dataset designed to aid in the detection and diagnosis of faults within deep learning systems. This dataset comprises over 14,000 training runs generated from 59 adapted TensorFlow/Keras programs, including nearly 10,000 faulty variants created through 27 source-code transformations. Each run is characterized by 26 features measuring training behavior at every epoch, such as weights, gradients, and accuracy trends. Deep4ge is intended to support tasks like binary fault detection, multi-class fault diagnosis, and early fault prediction from partial training data. AI
IMPACT Provides a standardized dataset for improving the reliability and robustness of deep learning models.
RANK_REASON The cluster contains an academic paper detailing a new dataset and benchmark for deep learning fault detection. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
