Pre-Training Meta-Rule Selection Policy for Visual Generative Abductive Learning

TL;DR

This paper tackles the efficiency challenge of logical abduction in visual generative abductive learning (AbdGen) when the meta-rule space is large. It introduces a pre-training-based meta-rule selection policy that leverages embeddings of symbol groundings and meta-rules, trained on pure symbolic data to keep costs low. The approach uses attention-based neural components and PPO-based policy optimization to shrink the meta-rule search space during AbdGen training, enabling faster and more scalable abductive learning. Empirical results on Mario and MNIST tasks show reduced training time and robust performance, even on unseen patterns and when grounding errors occur, with code available for reproduction.

Abstract

Visual generative abductive learning studies jointly training symbol-grounded neural visual generator and inducing logic rules from data, such that after learning, the visual generation process is guided by the induced logic rules. A major challenge for this task is to reduce the time cost of logic abduction during learning, an essential step when the logic symbol set is large and the logic rule to induce is complicated. To address this challenge, we propose a pre-training method for obtaining meta-rule selection policy for the recently proposed visual generative learning approach AbdGen [Peng et al., 2023], aiming at significantly reducing the candidate meta-rule set and pruning the search space. The selection model is built based on the embedding representation of both symbol grounding of cases and meta-rules, which can be effectively integrated with both neural model and logic reasoning system. The pre-training process is done on pure symbol data, not involving symbol grounding learning of raw visual inputs, making the entire learning process low-cost. An additional interesting observation is that the selection policy can rectify symbol grounding errors unseen during pre-training, which is resulted from the memorization ability of attention mechanism and the relative stability of symbolic patterns. Experimental results show that our method is able to effectively address the meta-rule selection problem for visual abduction, boosting the efficiency of visual generative abductive learning. Code is available at https://github.com/future-item/metarule-select.

Figures (8)

• Figure 1: Example of symbol-grounded conditional generation. Given the first picture and the symbolic rule Mario moves with right priority followed by up priority towards the target, a sequence of images would be generated. Here the symbol to be grounded for each image is the position of Mario, and the sub-symbolic style includes all visual features unchanged among images.
• Figure 2: The pre-training phase for learning the meta-rule selection policy.
• Figure 3: The application phase for applying the policy in AbdGen training.
• Figure 4: The changing probabilities of different meta-rules being selected for different tasks during pre-training.
• Figure 5: Performance on train tasks. Comparisons on train tasks Mario Right priority and MNIST Cumulative product of our meta-rule selection model, hand-made optimized meta-rules and random meta-rules on grounding accuracy ($\uparrow$) and training time($\uparrow$) over AbdGen iterations.