Various Graph Neural Networks (GNN) have been successful in analyzing data in non-Euclidean spaces, however, they have limitations such as oversmoothing, i.e., information becomes excessively averaged as the number of hidden layers increases. The issue stems from the intrinsic formulation of conventional graph convolution where the nodal features are aggregated from a direct neighborhood per layer across the entire nodes in the graph. As setting different number of hidden layers per node is infeasible, recent works leverage a diffusion kernel to redefine the graph structure and incorporate information from farther nodes. Unfortunately, such approaches suffer from heavy diagonalization of a graph Laplacian or learning a large transform matrix. In this regards, we propose a diffusion learning framework where the range of feature aggregation is controlled by the scale of a diffusion kernel. For efficient computation, we derive closed-form derivatives of approximations of the graph convolution with respect to the scale, so that node-wise range can be adaptively learned.With a downstream classifier, the entire framework is made trainable in an end-to-end manner. Our model is tested on various standard datasets for node-wise classification for the state-of-the-art performance, and it is also validated on a real-world brain network data for graph classifications to demonstrate its practicality for Alzheimer classification.

This study introduces a partially reconfigurable system-on-chip (SoC) platform leveraging dynamic resource management facilitated by a dynamic reconfigurable control processor (DRCP). By addressing the inherent reconfiguration time overheads of reconfigurable SoCs, the study demonstrates performance improvements through a runtime resource management strategy. The introduced management scheme effectively reduces the frequency of reconfigurations, thus lessening the associated overheads and increasing the operational efficiency of the SoC platform, which is designed to support on-chip multi-tenancy. Utilizing DRCP for dedicated resource management, the proposed SoC platform exhibited substantial reductions in reconfiguration times. When the partially reconfigurable SoC platform employed four partial regions (PRs), the reconfiguration counts decreased by 37.6%. Furthermore, upon extending the PRs to eight, there was a notable reduction in reconfiguration counts, achieving a decrease of 47.0%.

The current golden standard of semi-supervised semantic segmentation is to generate and exploit pseudo-supervision on unlabeled images. This approach is however susceptible to the quality of pseudo-supervision—training often becomes unstable particularly at early stages and biased to incorrect supervision. To address these issues, we propose a new semi-supervised learning framework, dubbed Guided Pseudo Supervision (GPS). GPS comprises three networks, i.e., a teacher and two separate students. The teacher is first trained with a small set of labeled data and provides stable initial pseudo-supervision on the unlabeled data to the students. The students interactively train each other under the supervision of the teacher, and once they are sufficiently trained, they offer feedback supervision to the teacher so that the teacher improves in subsequent iterations. This strategy enables more stable and faster convergence than previous works, and consequently, GPS achieved state-of-the-art performance on Pascal VOC 2012 and Cityscapes datasets in various experiment settings.

In analog-mixed-signal (AMS) compute-in-memory (CIM) systems, the two’s-complement (2SC) format provides better area efficiency than the sign-and-magnitude (SNM) one. However, the 2SC format exacerbates the challenges of AMS-CIM systems, suffering from significant DNN accuracy drop under process variations and high computation currents from activating multiple WLs. In the 2SC format, ‘0’ and ‘1’ are nearly balanced for all logical-order bits, unlike ‘0’-skewed higher-order bits in the SNM format. Consequently, the 2SC-based AMS-CIM systems have much more on-cells than the SNM-based counterpart, deteriorating the above challenges. We propose Skew-CIM, a software-hardware co-design technique to relax these challenges. Our proposed weight skewing (WESK) breaks the ‘0’ and ‘1’ balance at the software level. The potential accuracy drops resulting from WESK are successfully compensated by retraining DNNs. The offsets caused by WESK can be easily corrected using online hardware-level processing. Our Skew-CIM technique can be applied to most AMS-CIM systems with memories showing large on-off cell current ratios. As an example, we use it in a custom-designed 8T-SRAM-based CIM device, demonstrating a significant reduction in the DNN classification error by 7.6 times compared to the 2SC-based AMS-CIM without our Skew-CIM technique. Furthermore, our Skew-CIM markedly enhances energy efficiency by up to 39.9%, outperforming conventional SNM-based AMS-CIM systems.

This paper presents the fully-overlapped non-binary low-density parity-check (NB-LDPC) decoder to improve the throughput performance. The early-bubble (e-bubble) check algorithm and the EVN overlap scheduling is proposed to reduce the iteration latency of the decoder. The proposed decoder for (160, 80) NB-LDPC code over GF(64) achieved a throughput of 2.22 Gbps at a 625-MHz frequency.

The object-oriented functional data language (O/sup 2/FDL) is an interactive strongly typed database programming language that integrates the object-oriented and functional programming paradigms. It was designed for advanced applications that require a powerful and uniform database programming language supporting the software engineering principles of these tow paradigms. To this end, the O/sup 2/FDL supports inheritance and encapsulation of object-oriented languages, and an equational notation and strong typing of functional programming. Strong typing is extended with type restrictions to constrain the instantiations of type variables in function interfaces, while inheritance is augmented with a monotonic subtype rule for function redefinition. The O/sup 2/FDL also supports novel features. For the core of the O/sup 2/FDL, a denotational semantics based on an extended lambda -calculus is provided. For selected system-defined functions, operational semantics are provided and it is demonstrated that a subset of the O/sup 2/FDL is at least as expressive as the relational algebra.>

The Object-Oriented Functional Data Language (O/sup 2/FDL), an interactive, strongly typed database programming language, is discussed. It features inheritance and encapsulation of object-oriented languages and a functional notation for message expressions. The main contributions of the O/sup 2/FDL are: (1) functions of indefinite nesting levels that simplify nested list processing and form the basis of an efficient computation model; (2) path expressions that allow a flexible, graphical correspondence for database queries; (3) a type system featuring parametric and inclusion polymorphism, function restrictions, and type inferencing; and (4) system-defined functions that provide a concise and convenient notation for filtering aggregating, and combining objects. The versatility of the O/sup 2/FDL is demonstrated here for both relationally-complete expressions and more general computations.>