LIMEFLDL: A Local Interpretable Model-Agnostic Explanations Approach for Label Distribution Learning

Label distribution learning (LDL) is a novel machine learning paradigm that can handle label ambiguity. This paper focuses on the interpretability issue of label distribution learning. Existing local interpretability models are mainly designed for single-label learning problems and are difficult to directly interpret label distribution learning models. In response to this situation, we propose an improved local interpretable model-agnostic explanations algorithm that can effectively interpret any black-box model in label distribution learning.To address the label dependency problem, we introduce the feature attribution distribution matrix and derive the solution formula for explanations under the label distribution form. Meanwhile, to enhance the transparency and trustworthiness of the explanation algorithm, we provide an analytical solution and derive the boundary conditions for explanation convergence and stability. In addition, we design a feature selection scoring function and a fidelity metric for the explanation task of label distribution learning. A series of numerical experiments and human experiments were conducted to validate the performance of the proposed algorithm in practical applications. The experimental results demonstrate that the proposed algorithm achieves high fidelity, consistency, and trustworthiness in explaining LDL models.

Most AI models (like image classifiers) work by slapping simple labels on things. For example, calling a picture either "cat'' or "dog''. But real life isn’t that clear-cut. Sometimes an image might look 60% like a cat and 40% like a dog, that’s what we call label ambiguity. Traditional AI explanation tools only handle clear labels (like "100% cat") and can’t explain these "fuzzy" judgments. So our research tackles a key question: When an AI model says something is "60% cat, 40% dog'', which features (like ear shape or fur color) lead it to that conclusion? To solve this, we built an explainer that works with any black-box AI model handling label ambiguity. Specifically, we created a feature contribution matrix that measures how much each detail affects the fuzzy labels (e.g., "pointy ears boost the cat score by 30%''). We theoretically guaranteed the explanations are reliable and don’t contradict themselves. We designed tools to rank feature importance and check explanation accuracy, keeping things both simple and trustworthy. After testing with tons of data, our method outperforms existing tools. Our method explains fuzzy-label decisions more clearly and consistently, helping people actually understand and trust the AI’s reasoning.