ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) · 2017 · 113 citations · 0 references
ComputingComputer ScienceVisualizationComputer VisionData VisualizationMachine LearningDeep LearningMedical Image AnalysisMedical ImagingNeural NetworksRepresentation LearningDeep Learning-based ClassificationDifference VisualizationMedical Image ClassificationNeural Network Interpretability
Interpretable visualizations of neural network decisions are essential for improving models and enabling the use of black‑box classifiers in domains such as medicine. The study introduces prediction difference analysis as a method to visualize how a deep neural network responds to a specific input. The method computes the change in prediction probability when perturbing input pixels, highlighting evidence for or against a class. The technique successfully identifies image regions that support or contradict a class, surpassing previous visualization methods and providing deeper insight into classifier decision making.
This article presents the prediction difference analysis method for visualizing the response of a deep neural network to a specific input. When classifying images, the method highlights areas in a given input image that provide evidence for or against a certain class. It overcomes several shortcoming of previous methods and provides great additional insight into the decision making process of classifiers. Making neural network decisions interpretable through visualization is important both to improve models and to accelerate the adoption of black-box classifiers in application areas such as medicine. We illustrate the method in experiments on natural images (ImageNet data), as well as medical images (MRI brain scans).