The shortcomings of the Standard Model (SM) motivate its extension to accommodate new expected phenomena, such as dark matter and neutrino masses. However, such extensions are generally more complex due to the presence of a large number of free parameters and additional phenomenology. Understanding how theoretical and experimental limits affect the parameter spaces of new models, individually and collectively, is of utmost importance for conducting model status analysis, motivating precise computations, or model-building aimed at solving certain issues. However, checking the constraints usually require a large amount of time using a chain of physics tools. We demonstrate, for the first time, the application of deep learning (DL) for the multilabel classification (MLC) of a group of theoretical and experimental constraints in the dark doublet phase of the next-to-two-Higgs-doublet model (DDP-N2HDM), as a representative 9-dimensional parameter space. We analyze the issue of class imbalance and the ability of the classifier to learn joint class distributions. We demonstrate the time advantage compared to physics tools, with the classifier achieving orders of magnitude faster checks on groups of constraints and strong performance. The classifier performed strongly in terms of identifying regions where all constraints are valid or invalid, as well as regions where one or more of the constraints are valid or invalid simultaneously. This approach can be applied to any extension beyond the SM with the potential to aid HEP tools or act as a surrogate for fast model status checks. To that end, we provide a python tool \texttt{HEPMLC} for generating and investigating multilabel classifiers for SM extensions.
Mauricio A. Diaz, Srinandan Dasmahapatra, Stefano Moretti
This paper presents hep-aid, a modular Python library conceived for utilising, implementing, and developing parameter scan algorithms. Originally devised for sample-efficient, multi-objective active search approaches in computationally expensive Beyond Standard Model (BSM) phenomenology, the library currently integrates three Machine Learning (ML)-based approaches: a Constraint Active Search (CAS) algorithm, a multi-objective Active Search (AS) method (called b-CASTOR), and a self-exploration method named Machine Learning Scan (MLScan). These approaches address the challenge of multi-objective optimisation in high-dimensional BSM scenarios by employing surrogate models and strategically exploring parameter spaces to identify regions that satisfy complex objectives with fewer evaluations. Additionally, a Markov-Chain Monte Carlo method using the Metropolis-Hastings algorithm (MCMC-MH) is implemented for method comparison. The library also includes a High Energy Physics (HEP) module based on SPheno as the spectrum calculator. However, the library modules and functionalities are designed to be easily extended and used also with other external software for phenomenology. This manual provides an introduction on how to use the main functionalities of hep-aid and describes the design and structure of the library. Demonstrations based on the aforementioned parameter scan methods show that hep-aid methodologies enhance the efficiency of BSM studies, offering a versatile toolset for complex, multi-objective searches for new physics in HEP contexts exploiting advanced ML-based approaches.
Identifying disinfection byproducts (DBPs) with high health risk is an unresolved challenge. In this study, six members of a new class of aromatic nitrogenous DBPs─2-chloroaniline, 2-bromoaniline, 2,4-dichloroaniline, 2-chloro-4-bromoaniline, 4-chloro-3-nitroaniline, and 2-chloro-4-nitroaniline─are reported as DBPs in drinking water for the first time. Haloanilines completely degraded within 1 h in the presence of chlorine (1 mg/L), while about 20% remained in the presence of chloramine (1 mg/L) after 120 h. Haloanilines showed high stability in the absence of disinfectants, with <30% degradation at pH 5-9 over 120 h. Eight haloanilines were determined in chloraminated finished water and tap water at total concentrations of up to 443 ng/L. The most abundant was 2-bromoaniline, with a median concentration of 104 ng/L. The cytotoxicity of eight haloanilines and regulated trichloromethane and dichloroacetic acid (DCAA) was evaluated using Hep G2 cell assay. The EC50 values of eight haloanilines were 1-2 orders of magnitude lower than those of the regulated DBPs. The lowest toxic concentration of 2-chloro-4-nitroaniline was 1 μM, 500 times lower than that of DCAA. The formation and control of haloanilines in drinking water warrant further investigation.
Hericium erinaceus polysaccharides (HEPs) have attracted widespread attention in regulating gut microbiota (GM). To investigate digestibility and fermentation of HEPs and their effects on GM composition, three polysaccharide fractions, namely, HEP-30, HEP-50, and HEP-70, were fractionally precipitated with 30%, 50%, and 70% ethanol concentrations (v/v) from hot water-soluble extracts of Hericium erinaceus, respectively. Three kinds of prepared HEPs were structurally characterized and simulated gastrointestinal digestion, and their effects on human fecal microbiota fermentations of male and female and short-chain fatty acid (SCFA) production in vitro were clarified. Under digestive conditions simulating saliva, stomach, and small intestine, HEPs were not significantly influenced and safely reached the distal intestine. After 24 h of in vitro fermentation, the content of SCFAs was significantly enhanced (p < 0.05), and the retention rates of total and reducing sugars and pH value were significantly decreased (p < 0.05). Thus, HEPs could be utilized by GM, especially HEP-50, and enhanced the relative abundance of SCFA-producing bacteria, e.g., Bifidobacterium, Faecalibacterium, Blautia, Butyricicoccus, and Lactobacillus. Furthermore, HEPs reduced the relative abundances of opportunistic pathogenic bacteria, e.g., Escherichia-Shigella, Klebsiella, and Enterobacter. This study suggests that gradual ethanol precipitation is available for the preparation of polysaccharides from Hericium erinaceus, and the extracted polysaccharide could be developed as functional foods with great development value.
In this study, the digestion and fermentation properties of the bioactive water-soluble polysaccharide (HEP-W), and alkali-soluble polysaccharide (HEP-A) from Hericium erinaceus and the impact on the human colonic microbiota were determined using simulated saliva–gastrointestinal digestion and human fecal fermentation models in vitro. The basic physicochemical properties of HEP-W and HEP-A were determined at the same time. The results showed that the in vitro simulated digestion had almost no effect on the physicochemical properties of HEP-W and HEP-A, indicating that HEP-W and HEP-A were partially degraded. During fermentation, HEP-W and HEP-A increased the relative abundance of the dominant butyric acid-producing genera, the microbial community structure was significantly regulated, the gas production and short-chain fatty acid production in the fermentation broth were significantly increased, and the pH of the fermentation broth was reduced. There were structural and other differences in HEP-W and HEP-A due to different extraction methods, which resulted in different results. These results suggest that HEP-W and HEP-A may be potential gut microbial manipulators to promote gut health by promoting the production of beneficial metabolites by intestinal microorganisms using different butyric acid production pathways.
We present a machine learning approach for model-independent new physics searches. The corresponding algorithm is powered by recent large-scale implementations of kernel methods, nonparametric learning algorithms that can approximate any continuous function given enough data. Based on the original proposal by D’Agnolo and Wulzer (Phys Rev D 99(1):015014, 2019, arXiv:1806.02350 [hep-ph]), the model evaluates the compatibility between experimental data and a reference model, by implementing a hypothesis testing procedure based on the likelihood ratio. Model-independence is enforced by avoiding any prior assumption about the presence or shape of new physics components in the measurements. We show that our approach has dramatic advantages compared to neural network implementations in terms of training times and computational resources, while maintaining comparable performances. In particular, we conduct our tests on higher dimensional datasets, a step forward with respect to previous studies.