ApproxGNN: A Pretrained GNN for Parameter Prediction in Design Space Exploration for Approximate Computing

TL;DR

ApproxGNN introduces a pre-trained graph neural network model that predicts the quality and hardware cost of approximate accelerators, significantly improving prediction accuracy and transferability in design space exploration for approximate computing.

Contribution

It presents a novel transferable GNN-based approach with learned component embeddings for accurate QoR and HW cost prediction without retraining for each circuit.

Findings

50% improvement in mean square error over conventional methods

30% better overall prediction accuracy than non-finetuned ML approaches

54% improvement with fast finetuning

Abstract

Approximate computing offers promising energy efficiency benefits for error-tolerant applications, but discovering optimal approximations requires extensive design space exploration (DSE). Predicting the accuracy of circuits composed of approximate components without performing complete synthesis remains a challenging problem. Current machine learning approaches used to automate this task require retraining for each new circuit configuration, making them computationally expensive and time-consuming. This paper presents ApproxGNN, a construction methodology for a pre-trained graph neural network model predicting QoR and HW cost of approximate accelerators employing approximate adders from a library. This approach is applicable in DSE for assignment of approximate components to operations in accelerator. Our approach introduces novel component feature extraction based on learned embeddings rather than traditional error metrics, enabling improved transferability to unseen circuits. ApproxGNN models can be trained with a small number of approximate components, supports transfer to multiple prediction tasks, utilizes precomputed embeddings for efficiency, and significantly improves accuracy of the prediction of approximation error. On a set of image convolutional filters, our experimental results demonstrate that the proposed embeddings improve prediction accuracy (mean square error) by 50% compared to conventional methods. Furthermore, the overall prediction accuracy is 30% better than statistical machine learning approaches without fine-tuning and 54% better with fast finetuning.