MultiDelete for Multimodal Machine Unlearning

Jiali Cheng*, Hadi Amiri ;

"Machine Unlearning removes specific knowledge about training data samples from an already trained model. It has significant practical benefits, such as purging private, inaccurate, or outdated information from trained models without the need for complete re-training. Unlearning within a multimodal setting presents unique challenges due to the complex dependencies between different data modalities and the expensive cost of training on large multimodal datasets and architectures. This paper presents the first machine unlearning approach for multimodal data and models, titled , which is designed to decouple associations between unimodal data points during unlearning without losing the overall representation strength of the trained model. advocates for three key properties for effective multimodal unlearning: (a): , which effectively decouples the association between individual unimodal data points marked for deletion, rendering them as unrelated data points, (b): , which retains the multimodal representation post-unlearning, and (c): , which retains the unimodal representation post-unlearning. is efficient to train and is not constrained by using a strongly convex loss–a common restriction among existing baselines. Experiments on two architectures and four datasets, including image-text and graph-text datasets, show that gains an average improvement of 17.6 points over best performing baseline in unlearning multimodal samples, can maintain the multimodal and unimodal knowledge of the original model post unlearning, and can provide better protection to unlearned data against adversarial attacks1 . 1 Code and data is available at https://github.com/CLU-UML/MultiDelete"