Elevated osteoclast (OC) activity is a major contributor to inflammatory bone loss (IBL) during chronic inflammatory diseases. However, the specific OC precursors (OCPs) responding to inflammatory cues and the underlying mechanisms leading to IBL are poorly understood. We identified two distinct OCP subsets: Ly6C hi CD11b hi inflammatory OCPs (iOCPs) induced during chronic inflammation, and homeostatic Ly6C hi CD11b lo OCPs (hOCPs) which remained unchanged. Functional and proteomic characterization revealed that while iOCPs were rare and displayed low osteoclastogenic potential under normal conditions, they expanded during chronic inflammation and generated OCs with enhanced activity. In contrast, hOCPs were abundant and manifested high osteoclastogenic potential under normal conditions but generated OCs with low activity and were unresponsive to the inflammatory environment. Osteoclasts derived from iOCPs expressed higher levels of resorptive and metabolic proteins than those generated from hOCPs, highlighting that different osteoclast populations are formed by distinct precursors. We further identified the TNF-α and S100A8/A9 proteins as key regulators that control the iOCP response during chronic inflammation. Furthermore, we demonstrated that the response of iOCPs but not that of hOCPs was abrogated in tnf-α −/− mice, in correlation with attenuated IBL. Our findings suggest a central role for iOCPs in IBL induction. iOCPs can serve as potential biomarkers for IBL detection and possibly as new therapeutic targets to combat IBL in a wide range of inflammatory conditions.
Imperfections in data annotation, known as label noise, are detrimental tothe training of machine learning models and have an often-overlookedconfounding effect on the assessment of model performance. Nevertheless,employing experts to remove label noise by fully re-annotating large datasetsis infeasible in resource-constrained settings, such as healthcare. This workadvocates for a data-driven approach to prioritising samples for re-annotation- which we term "active label cleaning". We propose to rank instances accordingto estimated label correctness and labelling difficulty of each sample, andintroduce a simulation framework to evaluate relabelling efficacy. Ourexperiments on natural images and on a new medical imaging benchmark show thatcleaning noisy labels mitigates their negative impact on model training,evaluation, and selection. Crucially, the proposed active label cleaningenables correcting labels up to 4 times more effectively than typical randomselection in realistic conditions, making better use of experts' valuable timefor improving dataset quality.
Economic inequality is associated with preferences for smaller, immediate gains over larger, delayed ones. Such temporal discounting may feed into rising global inequality, yet it is unclear whether it is a function of choice preferences or norms, or rather the absence of sufficient resources for immediate needs. It is also not clear whether these reflect true differences in choice patterns between income groups. We tested temporal discounting and five intertemporal choice anomalies using local currencies and value standards in 61 countries (N = 13,629). Across a diverse sample, we found consistent, robust rates of choice anomalies. Lower-income groups were not significantly different, but economic inequality and broader financial circumstances were clearly correlated with population choice patterns.
Inspired by the developments in quantum computing, building domain-specificclassical hardware to solve computationally hard problems has receivedincreasing attention. Here, by introducing systematic sparsificationtechniques, we demonstrate a massively parallel architecture: the sparse IsingMachine (sIM). Exploiting sparsity, sIM achieves ideal parallelism: its keyfigure of merit - flips per second - scales linearly with the number ofprobabilistic bits (p-bit) in the system. This makes sIM up to 6 orders ofmagnitude faster than a CPU implementing standard Gibbs sampling. Compared tooptimized implementations in TPUs and GPUs, sIM delivers 5-18x speedup insampling. In benchmark problems such as integer factorization, sIM can reliablyfactor semiprimes up to 32-bits, far larger than previous attempts from D-Waveand other probabilistic solvers. Strikingly, sIM beats competition-winning SATsolvers (by 4-700x in runtime to reach 95% accuracy) in solving 3SAT problems.Even when sampling is made inexact using faster clocks, sIM can find thecorrect ground state with further speedup. The problem encoding andsparsification techniques we introduce can be applied to other Ising Machines(classical and quantum) and the architecture we present can be used for scalingthe demonstrated 5,000-10,000 p-bits to 1,000,000 or more through analog CMOSor nanodevices.
Massively parallel annealing processors may offer superior performance for awide range of sampling and optimization problems. A key component dictating thesize of these processors is the neuron update circuit, ideally implementedusing special stochastic nanodevices. We leverage photon statistics usingsingle photon avalanche diodes (SPADs) and temporal filtering to generatestochastic states. This method is a powerful alternative offering uniquefeatures not currently seen in annealing processors: the ability tocontinuously control the computational temperature and the seamless extensionto the Potts model, a $n$-state generalization of the two-state Ising model.SPADs also offer a considerable practical advantage since they are readilymanufacturable in current standard CMOS processes. As a first step towardsrealizing a CMOS SPAD-based annealer, we have designed Ising and Potts modelsdriven by an array of discrete SPADs and show they accurately sample from theirtheoretical distributions.
This paper presents a reluctance‐based model considering the frequency‐dependent loss nature of the windings for the analysis of three‐phase five‐leg transformers under grid voltages with direct current (DC) bias. This is very important especially for proper determination of their harmonic current distortion and maximum loading capability (MLC) under DC‐biased grid voltage conditions. To figure out the developed model's validity under sinusoidal and DC‐biased grid voltage cases, it is comparatively analyzed with the model based on 2D finite element method (FEM). Thus, for the considered transformer type operated under DC bias, the excitation current's harmonic pollution, losses, and reactive power demand parameters are analyzed by using the developed model. Additionally, by regarding these performance parameters, the DC susceptibilities of the considered‐type transformer and the single‐phase shell‐type transformer are comparatively evaluated. Finally, for the studied grid voltage conditions, the effects of two important design considerations as (i) magnetic core material selection and (ii) legs’ cross‐sectional area sizing on the MLC are investigated. It is concluded from these investigations that under saturation conditions, the transformers, which have the core material with higher permeability or lower reluctance, draw higher excitation current, and have lower MLC ratio when compared to ones having the core material with lower permeability or higher reluctance. However, for unsaturated transformers, which work under DC bias, the case is the opposite to that in saturation conditions. On the other hand, under DC bias conditions, the effect of cross‐sectional area sizing on the MLC ratio is much more for the transformer with high permeable magnetic core material with regards to ones with low permeable magnetic core material.