Search

Federated learning (FL) commonly encourages the clients to perform multiple local updates before the global aggregation, thus avoiding frequent model exchanges and relieving the communication bottleneck between the server and clients. Though empirically effective, the negative impact of multiple local updates on the stability of FL is not thoroughly studied, which may result in a globally unstable and slow convergence. Based on sensitivity analysis, we define in this paper a local-update stability index for the general FL, as measured by the maximum inter-client model discrepancy after the multiple local updates that mainly stems from the data heterogeneity. It enables to determine how much the variation of client’s models with multiple local updates may influence the global model, and can also be linked with the convergence and generalization. We theoretically derive the proposed local-update stability for current state-of-the-art FL methods, providing possible insight to understanding their motivation and limitation from a new perspective of stability. For example, naively executing the parallel acceleration locally at clients would harm the local-update stability. Motivated by this, we then propose a novel accelerated yet stabilized FL algorithm (named FedANAG) based on the server- and client-level Nesterov accelerated gradient (NAG). In FedANAG, the global and local momenta are elaborately designed and alternatively updated, while the stability of local update is enhanced with help of the global momentum. We prove the convergence of FedANAG for strongly convex, general convex and non-convex settings. We then conduct evaluations on both the synthetic and real-world datasets to first validate our proposed local-update stability. The results further show that across various data heterogeneity and client participation ratios, FedANAG not only accelerates the global convergence by reducing the required number of communication rounds to a target accuracy, but converges to an eventually higher accuracy.

Recent learned image compression models surpass manually designed methods in rate-distortion performance by introducing nonlinear transforms and end-to-end optimization. However, there still lack quantitative measurements that efficiently evaluate the latent representations inferred by learned image compression models. To address this problem, we develop novel measurements on robustness and importance of the latent representations. We first propose an admissible range that can be efficiently estimated via gradient ascent and descent for establishing the empirical distribution of latent representations. Consequently, the in-distribution region within the admissible range is derived to measure the robustness and channel importance of latent representations of natural images. Visualization demonstrates the statistics of latent representations are significantly distinguishing in the properties of robustness and linearity within and outside the in-distribution region. To our best knowledge, this paper proposes the first statistically meaningful measurements for learned image compression and successfully applies the measurements in corruption alleviation during successive image compression and post-training pruning in a training-free fashion. Compared with existing methods, the shrunk in-distribution constraint derived from the in-distribution region achieves superior robustness and rate-distortion performance in successive compression. The channel importance allows post-training pruning to achieve comparable rate-distortion performance with a reduction of up to 60% entropy coding time.

This article presents a novel high-gain antenna-on-chip (AoC) solution for the sub-terahertz band using a standard semiconductor and packaging process. The novel AoC is realized by bonding on a usual AoC in a semiconductor technology with a microbump in a flip-chip technology. As a result, an additional radiation mode can be activated, leading to increased radiation efficiency and gain. Both usual and novel AoCs were designed in the dipole-type antenna configuration for the band of 300 GHz. The usual AoC was fabricated using a $0.13 \, \mu \text{m}$ silicon-germanium (SiGe) technology, while the novel AoC was fabricated using the same $0.13 \, \mu \text{m}$ SiGe technology and the standard flip-chip technology. The usual and novel AoCs have the same die area of $2 \times 2$ mm2. Compared with the usual AoC, the novel AoC has demonstrated the enhancement of all performance parameters. The radiation efficiency, peak realized gain, and impedance bandwidth have been improved significantly, rising from 15.9% to 41.1%, 1.91 to 8.67 dBi, and 15.6 to 39.6 GHz, respectively. Hence, it is believed that the novel AoC solution is particularly suitable for highly integrated sub-terahertz systems.

A wafer probe is an essential feed for an integrated antenna during a measurement. The wafer probe radiates from its unshielded tips, which limits the minimum measurable gain of an antenna under test (AUT). Hence, the wafer probe radiation should be characterized, which, however, still remains to be attempted. This study aims to characterize it with the novel radiation standard structure (RSS) kit and by the radiation addition principle (RAP). More importantly, the characterized wafer probe is used to determine the radiation pattern of an integrated antenna. The errors in the radiation pattern due to the radiation and reflection from the wafer probe are removed by the radiation subtraction principle (RSP) and with the mathematical absorber reflection suppression (MARS) algorithm, respectively. The approaches are validated by simulations and experiments with an ACP65-A-GSG-500 wafer probe at 28 GHz.

Previous studies have indicated a potential relationship between iron status and oral health outcomes, specifically periodontitis and dental caries. This study employed Mendelian randomization (MR) to investigate the causal effects of iron status on these oral health conditions. The focus of this study was on key iron biomarkers, namely serum iron, ferritin, transferrin saturation (TSAT), and total iron-binding capacity (TIBC).

It is promising to solve linear inverse problems by unfolding iterative algorithms (e.g., iterative shrinkage thresholding algorithm (ISTA)) as deep neural networks (DNNs) with learnable parameters. However, existing ISTA-based unfolded algorithms restrict the network architectures for iterative updates with the partial weight coupling structure to guarantee convergence. In this paper, we propose hybrid ISTA to unfold ISTA with both pre-computed and learned parameters by incorporating free-form DNNs (i.e., DNNs with arbitrary feasible and reasonable network architectures), while ensuring theoretical convergence. We first develop HCISTA to improve the efficiency and flexibility of classical ISTA (with pre-computed parameters) without compromising the convergence rate in theory. Furthermore, the DNN-based hybrid algorithm is generalized to popular variants of learned ISTA, dubbed HLISTA, to enable a free architecture of learned parameters with a guarantee of linear convergence. To our best knowledge, this paper is the first to provide a convergence-provable framework that enables free-form DNNs in ISTA-based unfolded algorithms. This framework is general to endow arbitrary DNNs for solving linear inverse problems with convergence guarantees. Extensive experiments demonstrate that hybrid ISTA can reduce the reconstruction error with an improved convergence rate in the tasks of sparse recovery and compressive sensing.

Within the electronic design automation (EDA) domain, artificial intelligence (AI)-driven solutions have emerged as formidable tools, yet they typically augment rather than redefine existing methodologies. These solutions often repurpose deep learning models from other domains, such as vision, text, and graph analytics, applying them to circuit design without tailoring to the unique complexities of electronic circuits. Such an “AI4EDA” approach falls short of achieving a holistic design synthesis and understanding, overlooking the intricate interplay of electrical, logical, and physical facets of circuit data. This study argues for a paradigm shift from AI4EDA towards AI-rooted EDA from the ground up, integrating AI at the core of the design process. Pivotal to this vision is the development of a multimodal circuit representation learning technique, poised to provide a comprehensive understanding by harmonizing and extracting insights from varied data sources, such as functional specifications, register-transfer level (RTL) designs, circuit netlists, and physical layouts. We champion the creation of large circuit models (LCMs) that are inherently multimodal, crafted to decode and express the rich semantics and structures of circuit data, thus fostering more resilient, efficient, and inventive design methodologies. Embracing this AI-rooted philosophy, we foresee a trajectory that transcends the current innovation plateau in EDA, igniting a profound “shift-left” in electronic design methodology. The envisioned advancements herald not just an evolution of existing EDA tools but a revolution, giving rise to novel instruments of design-tools that promise to radically enhance design productivity and inaugurate a new epoch where the optimization of circuit performance, power, and area (PPA) is achieved not incrementally, but through leaps that redefine the benchmarks of electronic systems’ capabilities.

Aerocyanidin and amycomicin are two antibiotics derived from long-chain acids with a rare epoxy isonitrile moiety, the complexity of which renders the total synthesis of these two natural products rather challenging. How this functionality is biosynthesized has also remained obscure. While the biosynthetic gene clusters for these compounds have been identified, both appear to be deficient in genes encoding enzymes seemingly necessary for the oxidative modifications observed in these antibiotics. Herein, the biosynthetic pathways of aerocyanidin and amycomicin are fully elucidated. They share a conserved pathway to isonitrile intermediates that involves a bifunctional thioesterase and a nonheme iron α-ketoglutarate-dependent enzyme. In both cases, the isonitrile intermediates are then loaded onto an acyl carrier protein (ACP) catalyzed by a ligase. The isonitrile-tethered ACP is subsequently processed by polyketide synthase(s) to undergo chain extension, thereby assembling a long-chain γ-hydroxy isonitrile acid skeleton. The epoxide is installed by the cupin domain-containing protein AecF to conclude the biosynthesis of aerocyanidin. In contrast, three P450 enzymes AmcB, AmcC, and AmcQ are involved in epoxidation and keto formation to finalize the biosynthesis of amycomicin. These results thus explain the sequence of oxidation events that result in the final structures of aerocyanidin and amycomicin as well as the biosynthesis of the key γ-hydroxy epoxy isonitrile functional group.

Dendrimers are a family of nano‐sized three‐dimensional polymers with unique dendritic branching structures and compact spherical geometries. In recent years, dendrimers have made a series of breakthroughs in the biomedical field. In this review, we introduce the synthesis principles, modification methods, and new materials designed based on dendrimers; discuss the importance of cytotoxicity of dendrimers for applications; and elaborate on their applications in the field of molecular assembly and cancer diagnosis and treatment. We speculate that in the near future, more new materials based on dendrimers will be applied in the biomedical field. Dendrimers are a family of nano‐sized three‐dimensional polymers which now have attracted more extensive attention in biomedical field. This review provides insights into their synthesis methods, their applications in the field of biomedicine and their applications in cancer diagnosis and treatment. The limitations of dendrimers and the further development prospects are also discussed.

Hepatocellular carcinoma (HCC) is a hypervascular tumour characterized by tumour-driven neovascularization. The degrees of blood oxygen saturation (DBOS), microvessel density (MVD) and tumour size (TS) are indicators in identifying the development stage of HCC. Herein, we proposed an HCC staging model using HepG2 tumour-bearing mice based on DBOS, MVD and TS. According to the patterns of these three criteria, HCC was classified into four stages: early, intermediate, advanced and end stages. The advanced stage was characterized by MVD of 50–90 (number per mm2), DBOS of 12–16% and TS of 250–600 mm3, which poses a critical challenge in HCC therapy. In order to efficiently control and treat HCC in the advanced stage, we developed a cyclodextrin (CD)-based chaperoned inclusion complex using Sorafenib (Sor), β-CD and γ-CD (SCD) viathe co-crystallization method. The structural study manifested that CDs could encapsulate Sor with the hydrophobic cavities at a 1 : 1 stoichiometry ratio. The crystallographic analysis indicated that Sor-β-CD presented a diagonal stacking pattern, while Sor-γ-CD possessed a channel-type structure. The resultant chaperoned inclusion complexes significantly improved the solubility, dissolution rate and drug release of Sor, leading to superior pharmacokinetics, biodistribution and biosafety through oral administration. The antitumour effect was then evaluated on a mouse model with advanced HCC through oral administration and intratumour injection. The treatment involving the oral administration of SCDs showed a promising therapeutic effect on advanced HCC, which efficiently blocked angiogenesis and inhibited tumour progression. For the treatments using intratumour injections, only Sor-γ-CD exhibited a satisfactory anti-tumour effect with reduction in TS, MVD and DBOS. The enhanced therapeutic performance of Sor-γ-CD was attributed to its channel-type structure, which had an impact on the dissociation and release of the drug. Thus, Sor-γ-CD can be used as a potential pro-drug for clinical medicine and basic research to treat HCC.