Browse Articles
Discover research articles across all indexed journals
Multimodal MRI analysis of COVID-19 effects on pediatric brain
Molecular basis for the interaction between Saccharomyces cerevisiae Rtt103 and the Rat1-Rai1 complex
Low-profile conformal single-sided miniaturized frequency selective surface for wideband shielding
Optimization on multifractal loss landscapes explains a diverse range of geometrical and dynamical properties of deep learning
Abstract Gradient descent and its variants are foundational in solving optimization problems across many disciplines. In deep learning, these optimizers demonstrate a remarkable ability to dynamically navigate complex loss landscapes, ultimately converging to solutions that generalize well. To elucidate the mechanism underlying this ability, we introduce a theoretical framework that models the complexities of loss landscapes as multifractal. Our model unifies and explains a broad range of realistic geometrical signatures of loss landscapes, including clustered degenerate minima, multiscale structure, and rich optimization dynamics in deep neural networks, such as the edge of stability, non-stationary anomalous diffusion, and the extended edge of chaos without requiring fine-tuning parameters. We further develop a fractional diffusion theory to illustrate how these optimization dynamics, coupled with multifractal structure, effectively guide optimizers toward smooth solution spaces housing flatter minima, thus enhancing generalization. Our findings suggest that the complexities of loss landscapes do not hinder optimization; rather, they facilitate the process. This perspective not only has important implications for understanding deep learning but also extends potential applicability to other disciplines where optimization unfolds on complex landscapes.
Application of CT in predicting lymph node metastasis in cervical cancer and construction of a preoperative nomogram
Surface engineering on bulk Cu2O for efficient electrosynthesis of urea
Differences in symptom clusters based on multidimensional symptom experience and symptom burden in stroke patients
Motion-unrestricted dynamic electrocardiogram system utilizing imperceptible electronics
Twitter communities are associated with changing user’s opinion towards COVID-19 vaccine in Japan
Global structural survey of the flagellotropic myophage φTE infecting agricultural pathogen Pectobacterium atrosepticum
The health and wellbeing of undergraduate students in Indonesia: descriptive results of a survey in three public universities
Flexibility-tuning of dual-display DNA-encoded chemical libraries facilitates cyclic peptide ligand discovery
Abstract Cyclic peptides constitute an important drug modality since they offer significant advantages over small molecules and macromolecules. However, access to diverse chemical sets of cyclic peptides is difficult on a large library scale. DNA-encoded Chemical Libraries (DELs) provide a suitable tool to obtain large chemical diversity, but cyclic DELs made by standard DEL implementation cannot efficiently explore their conformational diversity. On the other hand, dual-display Encoded Self-Assembling Chemical (ESAC) Libraries can be used for modulating macrocycle flexibility since the two displayed peptides can be connected in an incremental fashion. In this work, we construct a 56 million dual-display ESAC library using a two-step cyclization strategy. We show that varying the level of conformational restraint is essential for the discovery of specific ligands for the three protein targets thrombin, human alkaline phosphatase and streptavidin.
Anatomical study of ultrasound vs landmark guidance for needle placement in the obliquus capitis inferior
Colonic epithelial-derived FGF1 drives intestinal stem cell commitment toward goblet cells to suppress inflammatory bowel disease
Compact modulators on silicon nitride waveguide platform via micro-transfer printing of thin-film lithium niobate
Abstract We explore the use of micro-transfer printing (µTP) technology to integrate thin lithium niobate (LN) films onto silicon nitride (SiN) waveguides, facilitating the development of compact electro-optical modulators. Three modulator architectures are investigated: Mach-Zehnder interferometer (MZI), Fabry-Perot (FP) resonator, and side-coupled FP resonators. By acting as a photonic molecule, the proposed coupled FP resonators enable improved spectral engineering with new functionalities while maximizing the transmission and quality-factor (Q-factor) of the resonances. Design, simulations, fabrication method, and experimental results are presented, demonstrating the potential of µTP in advancing electro-optical modulators. The half-wave voltage-length product (V π L) of the fabricated devices decreases as the Q-factor increases achieving V π L = 10.5, 4.3, and 2.74 V.cm for MZI, FP, and photonic molecule modulators, respectively.
Phase glides and self-organization of atomically abrupt interfaces out of stochastic disorder in α-Ga2O3
Abstract Disorder-induced ordering and remarkably high radiation tolerance in γ-phase of gallium oxide is a recent spectacular discovery at the intersection of the fundamental physics and electronic applications. Importantly, by far, these data were collected with initial samples in form of the thermodynamically stable β-phase of this material. Here, we investigate these phenomena starting from metastable α-phase and explain radically new trend occurring in the system. We argue that in contrast to that in β-to-γ disorder-induced transitions, the O sublattice in α-phase exhibits hexagonal close-packed structure, so that to activate α-to-γ transformation significant structural rearrangements are required in both Ga and O sublattices. Moreover, consistent with theoretical predictions, α-to-γ phase transformation requires accumulation of the substantial tensile strain to initiate otherwise impossible lattice glides. Thus, we explain the experimentally observed trends in term of the combination of disorder and strain governed process. Finally, we demonstrate atomically abrupt α/γ interfaces paradoxically self-organized out of the stochastic disorder.