Instance Space Analysis

Instance space analysis (ISA) focuses on understanding the properties of individual data points (instances) within a dataset to improve model performance, interpretability, and the generation of challenging benchmark problems. Current research employs techniques like Shapley values to quantify instance influence, and explores the use of genetic programming to generate synthetic datasets with specific characteristics. This work is significant for enhancing model explainability in areas such as medical image analysis and software defect prediction, as well as for developing more robust and challenging benchmarks for algorithm evaluation in optimization problems.