Simple Techniques Work Surprisingly Well for Neural Network Test Prioritization and Active Learning (Replicability Study)
This is an incremental replicability study for researchers and practitioners in machine learning testing, showing that basic uncertainty quantification methods are as effective as specialized techniques.
The study tackled the problem of efficiently prioritizing test inputs for deep neural networks to save computation and labeling costs, finding that simple techniques like predicted softmax likelihood or entropy perform equally well as the more complex DeepGini method in large-scale experiments.
Test Input Prioritizers (TIP) for Deep Neural Networks (DNN) are an important technique to handle the typically very large test datasets efficiently, saving computation and labeling costs. This is particularly true for large-scale, deployed systems, where inputs observed in production are recorded to serve as potential test or training data for the next versions of the system. Feng et. al. propose DeepGini, a very fast and simple TIP, and show that it outperforms more elaborate techniques such as neuron- and surprise coverage. In a large-scale study (4 case studies, 8 test datasets, 32'200 trained models) we verify their findings. However, we also find that other comparable or even simpler baselines from the field of uncertainty quantification, such as the predicted softmax likelihood or the entropy of the predicted softmax likelihoods perform equally well as DeepGini.