Graph Neural Networks (GNNs) have recently emerged as a robust framework for graph-structured data. They have been applied to many problems such as knowledge graph analysis, social networks recommendation, and even Covid19 detection and vaccine developments. However, unlike other deep neural networks such as Feed Forward Neural Networks (FFNNs), few analyses such as verification and property inferences exist, potentially due to dynamic behaviors of GNNs, which can take arbitrary graphs as input, whereas FFNNs which only take fixed size numerical vectors as inputs. This paper proposes an approach to analyze GNNs by converting them into FFNNs and reusing existing FFNNs analyses. We discuss various designs to ensure the scalability and accuracy of the conversions. We illustrate our method on a study case of node classification. We believe that our approach opens new research directions for understanding and analyzing GNNs.
Binary neural networks (BNNs) that use 1-bit weights and activations have garnered interest as extreme quantization provides low power dissipation. By implementing BNNs as computing-in-memory (CIM), which computes multiplication and accumulations on memory arrays in an analog fashion, namely analog CIM, we can further improve the energy efficiency to process neural networks. However, analog CIMs suffer from the potential problem that process variation degrades the accuracy of BNNs. Our Monte-Carlo simulations show that in an SRAM-based analog CIM of VGG-9, the classification accuracy of CIFAR-10 is degraded even below 20% under process variations of 65nm CMOS. To overcome this problem, we present a variation-aware BNN framework. The proposed framework is developed for SRAM-based BNN CIMs since SRAM is most widely used as on-chip memory, however easily extensible to BNN CIMs based on other memories. Our extensive experimental results show that under process variation of 65nm CMOS, our framework significantly improves the CIFAR-10 accuracies of SRAM-based BNN CIMs, from 10% and 10.1% to 87.76% and 77.74% for VGG-9 and RESNET-18 respectively.
Convolutional Neural Networks have demonstrated excellent ability in image recognition, yet they frequently have significant computational and memory requirements. It is advantageous to create a tensor factorization framework that can effectively compress networks in order to solve this problem. With its ability to express high-dimensional tensors using a smaller set of core tensors and fewer parameters, the Tensor-Train (TT) structure is a viable choice for compressing neural networks. The choice of appropriate TT-ranks, however, is now without theoretical guarantees. In our study, we introduce a method that employs the widely recognized heuristic network slimming structure through batch normalization techniques. This method serves as metrics to gauge the requisite parameter count, encompassing the intricacies of the training data complexity derived from the original CNNs. Our focus remains on the TT-rank, which denotes the scale parameter choice facilitated by the estimated size. This approach provides a comprehensive strategy to ascertain network compression requirements and resource optimization. Our results reveal that our calculated TT-ranks lead to substantial reduction in computational complexity and memory usage, while maintaining competitive accuracy compared to baseline CNNs.