Compare-xAI: Toward Unifying Functional Testing Methods for Post-hoc XAI Algorithms into a Multi-dimensional Benchmark.
Mohamed Karim BelaidRichard BornemannMaximilian RabusRalf KrestelEyke HüllermeierPublished in: xAI (2) (2023)
Keyphrases
- post hoc
- multi dimensional
- significant improvement
- machine learning methods
- computationally expensive
- computational cost
- computationally complex
- benchmark datasets
- computationally demanding
- methods outperform
- statistical methods
- theoretical guarantees
- problems in computer vision
- alternative methods
- heuristic methods
- learning algorithm
- experimental comparison
- synthetic and real datasets
- data sets
- computer vision algorithms
- algorithms require
- noisy data
- high computational complexity
- data streams
- methods can be applied
- preprocessing
- empirical studies
- times faster
- methods require
- computationally intensive
- parameter settings
- optimization methods
- quality measures
- complexity analysis
- exhaustive search
- hybrid method