MPRU: Modular Projection-Redistribution Unlearning as Output Filter for Classification Pipelines

2025-10-30

AI Summary

Existing machine unlearning (MU) methods rely on access to the original training data and the full model, resulting in poor scalability and high deployment costs. This paper proposes MPRU—a modular, model-agnostic output filter that achieves knowledge forgetting via plug-and-play post-processing, without requiring access to either the original training data or model parameters. Its core innovation lies in modeling classification training as a reversible sequential process and introducing a projection-redistribution layer that reconstructs decision boundaries directly in the output space. MPRU is compatible with diverse classifiers—including CNNs and tree-based models—and integrates seamlessly into existing ML pipelines. Experiments on image and tabular datasets demonstrate that MPRU matches the unlearning efficacy of full retraining while reducing computational overhead significantly. These results validate MPRU’s efficiency, generality, and scalability.

Technology Category

• Machine Learning: Privacy
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

As a new and promising approach, existing machine unlearning (MU) works typically emphasize theoretical formulations or optimization objectives to achieve knowledge removal. However, when deployed in real-world scenarios, such solutions typically face scalability issues and have to address practical requirements such as full access to original datasets and model. In contrast to the existing approaches, we regard classification training as a sequential process where classes are learned sequentially, which we call emph{inductive approach}. Unlearning can then be done by reversing the last training sequence. This is implemented by appending a projection-redistribution layer in the end of the model. Such an approach does not require full access to the original dataset or the model, addressing the challenges of existing methods. This enables modular and model-agnostic deployment as an output filter into existing classification pipelines with minimal alterations. We conducted multiple experiments across multiple datasets including image (CIFAR-10/100 using CNN-based model) and tabular datasets (Covertype using tree-based model). Experiment results show consistently similar output to a fully retrained model with a high computational cost reduction. This demonstrates the applicability, scalability, and system compatibility of our solution while maintaining the performance of the output in a more practical setting.

Problem

• Developing scalable machine unlearning without full dataset access
• Implementing modular output filters for classification pipelines
• Achieving efficient knowledge removal while maintaining model performance

Innovation

• Appends projection-redistribution layer to model
• Reverses last training sequence for unlearning
• Deploys modular output filter into pipelines