Gradient aware adaptive quantization: Locally uniform quantization with learnable clipping thresholds for globally non-uniform weights.

Non-uniform quantization has been shown to achieve promising performance for compressing neural networks, due to its better adaptation to the distribution of weights. However, traditional non-uniform quantization methods rely solely on weight distribution density, resulting in diminished model performance post-quantization. To tackle this challenge, we propose a novel non-uniform quantization method that can not only automatically learn the clipping threshold but also adaptively adjust the quantization levels, which can effectively reduce the quantization error.

Crystal-LSBO: Automated Design of De Novo Crystals With Latent Space Bayesian Optimization.

Generative modeling of crystal structures is significantly challenged by the complexity of input data, which constrains the ability of these models to explore and discover novel crystals. This complexity often confines de novo design methodologies to merely small perturbations of known crystals and hampers the effective application of advanced optimization techniques. One such optimization technique, latent space Bayesian optimization (LSBO), has demonstrated promising results in uncovering novel objects across various domains, especially when combined with variational autoencoders (VAEs).

Network structure influences self-organized criticality in neural networks with dynamical synapses.

The brain criticality hypothesis has been a central research topic in theoretical neuroscience for two decades. This hypothesis suggests that the brain operates near the critical point at the boundary between order and disorder, where it acquires its information-processing capabilities. The mechanism that maintains this critical state has been proposed as a feedback system known as self-organized criticality (SOC); brain parameters, such as synaptic plasticity, are regulated internally without external adjustment.

Pedology and Plant Provenance Can Improve Species Distribution Predictions of Australian Native Flora: A Calibrated and Validated Modeling Exercise on 5033 Species.

Species distribution models (SDMs) are valuable tools for assessing species' responses to environmental factors and identifying areas suitable for their survival. The careful selection of input variables is critical, as their interactions and correlations with other environmental factors can affect model performance. This study evaluates the influence of climate and soil variables on the performance of SDMs for 5033 Australian terrestrial vascular plant species, representing the largest phylogenetic diversity of native flora assessed in such an analysis.

Clinical and Economic Assessment of Robot-Assisted Radical Prostatectomy With the Hinotori Surgical System: Insights From a Japanese Single-Centre Study.

Background: We investigated the clinical and economic outcomes of robot-assisted radical prostatectomy (RARP) using the hinotori surgical system (H-RARP).

Methods: The analysis was conducted on 76 and 68 patients who underwent RARP using the da Vinci surgical system (D-RARP) and H-RARP, respectively. Safety, oncological, and functional outcomes were compared between H-RARP and D-RARP.

Optimization of crossing strategy based on the usefulness criterion in interpopulation crosses considering different marker effects among populations.

Herein, a method has been proposed for selecting optimal cross pairs based on the genetic potential of progeny in interpopulation crosses, considering different genetic effects among populations. In the breeding programs for self-pollinating plants, genetic improvements in multiple traits can be challenging when relying solely on a single biparental population, and interpopulation crosses are employed to integrate favorable alleles from multiple biparental populations to overcome this limitation. In this context, it is crucial to consider distinct genetic effects in different populations.

Polyunsaturated fatty acids in lipid membranes regulate human neuronal function and amyloid-β production.

The effects and mechanisms of polyunsaturated fatty acids (PUFAs) including docosahexaenoic acid (DHA) and arachidonic acid (ARA) contained in the lipid membrane of neurons in the production of amyloid β (Aβ), a pathogenic molecule in Alzheimer's disease (AD), remain unclear. In this study, we cultured human cortical neurons differentiated from induced pluripotent stem cells (iPSCs) under conditions of PUFA deficiency being progressively alleviated. Under PUFA-deficient conditions, increasing the total PUFA composition ratio in the lipid membrane enhanced membrane fluidity and reduced Aβ production.

Analyzing coordinated group behavior through role-sharing: a pilot study in female 3-on-3 basketball with practical application.

A group often shares a common goal and accomplishes a task that is difficult to complete alone by distributing roles. In such coordination, the non-verbal behavior among three or more members complicates the explanation of the mechanism due to complex and dynamic interactions. In cognitive science, a crucial role is indicated: to intervene moderately with others and adjust the whole balance without interrupting their main smooth interactions, using an experimental task.

Development of an electron spin resonance spectroscopy code for measuring carbonate radicals in tooth enamel and verification of its practicality using irradiated Japanese macaque teeth.

Purpose: We developed a new computer program for the application of electron spin resonance (ESR) to dosimetry of wild animals related to the Fukushima Daiichi Nuclear Power Plant accident.

Materials & Methods: The ESR spectra of carbonate radicals and other inorganic radicals are calculated by the complete elliptic integral. A simulated annealing method is implemented for the parameter optimization.

Implications From the Analogous Relationship Between Evolutionary and Learning Processes.

Organismal evolution is a process of discovering better-fitting phenotypes through trial and error across generations. This iterative process resembles learning processes, an analogy recognized since the 1950s. Recognizing this parallel suggests that evolutionary biology and machine learning can mutually benefit from each other; however, ample opportunities for research into their corresponding concepts remain.

Discovery and affinity maturation of antibody fragments from an unfavorably enriched phage display selection by deep sequencing and machine learning.

Phage display selection has been used for directed evolution of antibody fragments. However, variants with binding affinity cannot be always identified due to undesirable enrichment of target-unrelated variants in the biopanning process. Here, our goal was to obtain functional variants by deep sequencing and machine learning from a phage display library where functional variants were not appropriately enriched.

Neural basis of self-esteem: social cognitive and emotional regulation insights.

Introduction: Self-esteem (SE) can significantly affect individual well-being and has been linked to various psychiatric conditions. SE involves cognitive and emotional regulation within a social context. Prior research focusing on young adults has indicated neural correlations in prefrontal cortex areas but presented inconsistent findings.

Learning from small datasets-review of workshop 6 of the 10th International BCI Meeting 2023.

In a brain-computer interface (BCI), a primary objective is to reduce calibration time by recording as few as possible novel data points to (re-)train decoder models.Minimizing the calibration can be crucial for enhancing the usability of a BCI application with patients, increasing the acceptance by healthy users, facilitating a fast adaptation during non-stationary recordings, or transferring between sessions.At the 10th International BCI Meeting in 2023, our workshop addressed the latest proposed techniques to train classification or regression machine learning models with small datasets.

Allostery Links hACE2 Binding, Pan-variant Neutralization and Helical Extension in the SARS-CoV-2 Spike Protein.

The SARS-CoV-2 spike protein is highly antigenic, with epitopes in three distinct regions of the receptor binding domain (RBD) alone that have known mechanisms of neutralization. In previous work, we predicted a fourth RBD epitope based on allosteric conformational perturbations measured by hydrogen-deuterium exchange mass spectrometry (HDX-MS) upon complexation with the canonical spike protein target, human angiotensin-converting enzyme 2 (hACE2). We subsequently identified a pan-neutralizing antibody (ICO-hu104) with the predicted epitope, however, as the epitope was somewhat distant from the hACE2 binding interface, and our previous work limited to the spike RBD, the neutralization mechanism was unclear.

Deep-embedded clustering by relevant scales and genome-wide association study in autism.

Autism spectrum disorder (ASD) presents with heterogeneous phenotypic and genetic characteristics. Despite investigation into the molecular mechanisms underlying ASD, its etiology remains elusive. In our previous investigation within the Simons Simplex Collection (SSC), we noted increased signals through a genome-wide association study (GWAS) by clustering patients with ASD and reducing the sample size.

Panel Data Analysis of Socioeconomic Factors and COVID-19's Impact on Drinking Habits: Evidence from a Japanese Survey.

Alcohol use is closely related to health, emotional state, and social behavior. However, no clear consensus exists on how socioeconomic factors influence drinking habits or how the COVID-19 pandemic affected alcohol use. This study examines these relationships in Japan using panel data from a questionnaire survey (2014-2022), consisting of 10,836 responses provided by 1289 respondents, and logistic regression analysis.

Relationships Between Self-Esteem and Personal Attributes, Income, Consumption, and Assets: Japanese Panel Study.

Self-esteem is a key topic in psychology and health research. It influences well-being, happiness, and even medicine. However, existing studies on self-esteem have yielded conflicting results, suggesting that a global consensus remains elusive.

Respiratory syncytial virus infection, non-respiratory syncytial virus respiratory infections, and later wheezing.

Background: Studies investigating whether respiratory syncytial virus (RSV) infection, non-RSV respiratory infections, respiratory-related disorders, and non-respiratory-related disorders are associated with subsequent wheezing are limited in Japan. We aimed to elucidate the relationship between hospitalization for RSV infection, non-RSV respiratory infections, respiratory-related disorders, as well as non-respiratory-related disorders and subsequence wheezing in Japanese children.

Methods: This study included 7340 children and was conducted under the TMM BirThree Cohort Study (Tohoku Medical Megabank Project Birth and Three-Generation Cohort Study).

KCNN4-mediated potassium ion efflux maintains mitochondrial functions leading to platelet biogenesis.

Background: Potassium ions (K) are essential for platelet function, yet their role in thrombopoiesis-particularly through specific K channels-remains poorly understood. This gap is especially relevant in the context of in vitro platelet production from induced pluripotent stem cell (iPSC)-derived immortalized megakaryocyte progenitor cell lines (imMKCLs), which we developed for clinical-grade platelet manufacturing.

Objectives: We aimed to elucidate how K channels contribute to platelet biogenesis, focusing specifically on the calcium ion (Ca)-activated K channel KCNN4 (also known as K3.

Impact of germline variants on breast and ovarian cancer risk in Japanese women: an original cohort study and meta-analysis.

Background: Pathogenic variants (PVs) of BRCA1 and BRCA2 predispose individuals to a higher risk of breast and ovarian cancer; however, the precise risks posed by other cancer susceptibility genes remain unclear, particularly in Asian populations.

Methods: We executed a case-control study of 11 and 26 genes associated with breast and ovarian cancer susceptibility, respectively, in 7220 women with breast cancer, 2464 women with ovarian cancer, and 4032 controls from a multicentre, hospital-based registry in Japan. Furthermore, we conducted a meta-analysis of 23,193 patients with breast and/or ovarian cancer and 31,190 controls from six other hospital-based studies.

Real-World Experience with Triplet Therapy for High-Volume Metastatic Castration-Sensitive Prostate Cancer: A Retrospective Cohort Study from a Japanese Academic Hospital.

Background: Prostate cancer (PCa) significantly contributes to male cancer mortality. Triplet therapy shows promise for metastatic castration-sensitive prostate cancer (mCSPC), but real-world data are limited. This study aimed to evaluate the clinical outcomes of triplet therapy in real-world patients with mCSPC at an academic hospital in Japan.

SEgene identifies links between super enhancers and gene expression across cell types.

Enhancers are non-coding DNA regions that facilitate gene transcription, with a specialized subset, super-enhancers, known to exert exceptionally strong transcriptional activation effects. Super-enhancers have been implicated in oncogenesis, and their identification is achievable through histone mark chromatin immunoprecipitation followed by sequencing data using existing analytical tools. However, conventional super-enhancer detection methodologies often do not accurately reflect actual gene expression levels, and the large volume of identified super-enhancers complicates comprehensive analysis.

Kernel Bayesian tensor ring decomposition for multiway data recovery.

Tensor ring (TR) decomposition has emerged as the prevailing method for tensor completion. Earlier approaches have situated TR decomposition within a probabilistic framework, yielding satisfactory outcomes. However, these methods ignore side information or are inherently incapable of leveraging it.

HDXRank: A Deep Learning Framework for Ranking Protein Complex Predictions with Hydrogen-Deuterium Exchange Data.

Accurate modeling of protein-protein complex structures is essential for understanding biological mechanisms. Hydrogen-deuterium exchange (HDX) experiments provide valuable insights into binding interfaces. Incorporating HDX data into protein complex modeling workflows offers a promising approach to improve prediction accuracy.