Assessment of differentially private synthetic data for utility and fairness in end-to-end machine learning pipelines for tabular data.

Differentially private (DP) synthetic datasets are a solution for sharing data while preserving the privacy of individual data providers. Understanding the effects of utilizing DP synthetic data in end-to-end machine learning pipelines impacts areas such as health care and humanitarian action, where data is scarce and regulated by restrictive privacy laws. In this work, we investigate the extent to which synthetic data can replace real, tabular data in machine learning pipelines and identify the most effective synthetic data generation techniques for training and evaluating machine learning models. We systematically investigate the impacts of differentially private synthetic data on downstream classification tasks from the point of view of utility as well as fairness. Our analysis is comprehensive and includes representatives of the two main types of synthetic data generation algorithms: marginal-based and GAN-based. To the best of our knowledge, our work is the first that: (i) proposes a training and evaluation framework that does not assume that real data is available for testing the utility and fairness of machine learning models trained on synthetic data; (ii) presents the most extensive analysis of synthetic dataset generation algorithms in terms of utility and fairness when used for training machine learning models; and (iii) encompasses several different definitions of fairness. Our findings demonstrate that marginal-based synthetic data generators surpass GAN-based ones regarding model training utility for tabular data. Indeed, we show that models trained using data generated by marginal-based algorithms can exhibit similar utility to models trained using real data. Our analysis also reveals that the marginal-based synthetic data generated using AIM and MWEM PGM algorithms can train models that simultaneously achieve utility and fairness characteristics close to those obtained by models trained with real data.

Is ChatGPT a reliable source of scientific information regarding third-molar surgery?

BACKGROUND: ChatGPT (OpenAI) is a large language model. This model uses artificial intelligence and machine learning techniques to generate humanlike language and responses, even to complex questions. The authors aimed to assess the reliability of responses provided via ChatGPT and evaluate its trustworthiness as a means of obtaining information about third-molar surgery. METHODS: The authors assessed the 10 most frequently asked questions about mandibular third-molar extraction. A validated questionnaire (Chatbot Usability Questionnaire) was used and 2 oral and maxillofacial surgeons compared the answers provided with the literature. RESULTS: Most of the responses (90.63%) provided via the ChatGPT platform were considered safe and accurate and followed what was the stated in the English-language literature. CONCLUSIONS: The ChatGPT platform offers accurate and scientifically backed answers to inquiries about third-molar surgical extraction, making it a dependable and easy-to-use resource for both patients and the general public. However, the platform should provide references with the responses to validate the information. PRACTICAL IMPLICATIONS: Patients worldwide are exposed to reliable information sources. Oral surgeons and health care providers should always advise patients to be aware of the information source and that the ChatGPT platform offers a reliable solution.

Impacts of Farrowing Pen Design, Season, and Sow Parity on Litter Performance and Piglet Mortality.

Piglet mortality during lactation is a significant concern in swine production, influenced by complex interactions involving sow, piglet, environmental, and management factors. While crushing by the sow may be the ultimate cause of piglet mortality, there are many factors influencing the outcome, including parity, thermal stress, and animal housing systems. New farrowing systems are continuously being developed; however, it is difficult for producers to make decisions without any scientific basis. This study aimed to assess the impact of different farrowing pen layouts on piglet performance, considering parity and season. A total of 546 sows and 9123 piglets were monitored across 36 lactation cycles. Sows were randomly assigned to three farrowing pen layouts (standard, diagonal, and offset) in three rooms (20 sows by room). All farrowing pens had the same space allocations (2.7 m by 1.8 m and 2.1 m by 0.6 m for the sow area). The three types of farrowing pens were blocked by position within the room. Piglet performance traits (percent of stillborns, percent of mortality, percent of overlays, and average daily weight gain: ADG) and sows traits (health and parity) were monitored following US Meat Animal Research Center (USMARC) procedures. Results indicated that treatment, parity, and season influenced some piglet performance traits. The offset farrowing pen had a lower percent of stillborns compared to the standard. No significant differences were observed between the diagonal crate and the other treatments. When evaluating high mortality sow (>two piglets), the offset and standard treatments had a lower percent of overlays. Piglets from first-parity sows had lower ADG than those from higher-parity sows. A higher percent of overlays were observed in Autumn and Summer compared to Spring and Winter, and Summer had lower average daily weight gain than other seasons. The results suggest that modifying the layout (offset), with sows placed further away from the heating source, can reduce the percent of overlays in sows with high mortality (>2 piglets). In addition, the influence of season on the piglet production traits demonstrated the importance of proper management of the environment, even in systems with a certain level of climatic control.

Deep Learning Models to Predict Finishing Pig Weight Using Point Clouds.

The selection of animals to be marketed is largely completed by their visual assessment, solely relying on the skill level of the animal caretaker. Real-time monitoring of the weight of farm animals would provide important information for not only marketing, but also for the assessment of health and well-being issues. The objective of this study was to develop and evaluate a method based on 3D Convolutional Neural Network to predict weight from point clouds. Intel Real Sense D435 stereo depth camera placed at 2.7 m height was used to capture the 3D videos of a single finishing pig freely walking in a holding pen ranging in weight between 20-120 kg. The animal weight and 3D videos were collected from 249 Landrace × Large White pigs in farm facilities of the FZEA-USP (Faculty of Animal Science and Food Engineering, University of Sao Paulo) between 5 August and 9 November 2021. Point clouds were manually extracted from the recorded 3D video and applied for modeling. A total of 1186 point clouds were used for model training and validating using PointNet framework in Python with a 9:1 split and 112 randomly selected point clouds were reserved for testing. The volume between the body surface points and a constant plane resembling the ground was calculated and correlated with weight to make a comparison with results from the PointNet method. The coefficient of determination (R2 = 0.94) was achieved with PointNet regression model on test point clouds compared to the coefficient of determination (R2 = 0.76) achieved from the volume of the same animal. The validation RMSE of the model was 6.79 kg with a test RMSE of 6.88 kg. Further, to analyze model performance based on weight range the pigs were divided into three different weight ranges: below 55 kg, between 55 and 90 kg, and above 90 kg. For different weight groups, pigs weighing below 55 kg were best predicted with the model. The results clearly showed that 3D deep learning on point sets has a good potential for accurate weight prediction even with a limited training dataset. Therefore, this study confirms the usability of 3D deep learning on point sets for farm animals' weight prediction, while a larger data set needs to be used to ensure the most accurate predictions.

Acute hemodynamic responses from low-load resistance exercise with blood flow restriction in young and older individuals: A systematic review and meta-analysis of cross-over trials.

OBJECTIVE: To summarize the existing evidence on the acute response of low-load (LL) resistance exercise (RE) with blood flow restriction (BFR) on hemodynamic parameters. DATA SOURCES: MEDLINE (via PubMed), EMBASE (via Scopus), SPORTDiscus, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Web of Science and MedRxiv databases were searched from inception to February 2022. REVIEW METHODS: Cross-over trials investigating the acute effect of LLRE + BFR versus passive (no exercise) and active control methods (LLRE or HLRE) on heart rate (HR), systolic (SBP), diastolic (DBP) and mean (MBP) blood pressure responses. RESULTS: The quality of the studies was assessed using the PEDro scale, risk of bias using the RoB 2.0 tool for cross-over trials and certainty of the evidence using the GRADE method. A total of 15 randomized cross-over studies with 466 participants were eligible for analyses. Our data showed that LLRE + BFR increases all hemodynamic parameters compared to passive control, but not compared to conventional resistance exercise. Subgroup analysis did not demonstrate any differences between LLRE + BFR and low- (LL) or high-load (HL) resistance exercise protocols. Studies including younger volunteers presented higher chronotropic responses (HR) than those with older volunteers. CONCLUSIONS: Despite causing notable hemodynamic responses compared to no exercise, the short-term LL resistance exercise with BFR modulates all hemodynamic parameters HR, SBP, DBP and MBP, similarly to a conventional resistance exercise protocol, whether at low or high-intensity. The chronotropic response is slightly higher in younger healthy individuals despite the similarity regarding pressure parameters.

Ab initio study of the adsorption properties of CO2 reduction intermediates: The effect of Ni5Ga3 alloy and the Ni5Ga3/ZrO2 interface.

The Ni5Ga3 alloy supported on ZrO2 is a promising catalyst for the reduction of CO2 due to its higher selectivity to methanol at ambient pressure, e.g., activity comparable to industrial catalysts. However, our atomistic understanding of the role of the cooperative effects induced by the Ni5Ga3 alloy formation and its Ni5Ga3/ZrO2 interface in the CO2 reduction is still far from satisfactory. In this work, we tackle these questions by employing density functional theory calculations to investigate the adsorption properties of key CO2 reduction intermediates (CO2, H2, cis-COOH, trans-COOH, HCOO, CO, HCO, and COH) on Ni8, Ga8, Ni5Ga3, (ZrO2)16, and Ni5Ga3/(ZrO2)16. We found that Ni containing clusters tended to assume wetting configurations on the (ZrO2)16 cluster, while the presence of Ga atoms weakens the adsorption energies on the oxide surface. We also observed that CO2 was better activated on the metal-oxide interfaces and on the oxide surface, where it was able to form CO3-like structures. Meanwhile, H2 activation was only observed on Ni sites, which indicates the importance of distinct adsorption sites that can favor different CO2 reduction steps. Moreover, the formation of the metal-oxide interface showed to be beneficial for the adsorption of COOH isomers and unfavorable for the adsorption of HCOO.

The role of lithium treatment on comorbid anxiety symptoms in patients with bipolar depression.

BACKGROUND: Comorbid anxiety is pervasive and carries an immense psychosocial burden for patients with bipolar disorder. Despite this, trials reporting anxiety-related outcomes in this population are uncommon, particularly with regards to monotherapies. METHODS: Patients (n = 31) with both bipolar I or II disorder in current depressive episodes were enrolled in a six-week, open-label, single-center trial assessing the efficacy of lithium monotherapy in treating symptoms depression and comorbid anxiety. Patients were mostly medication-free and lithium-naïve at baseline. RESULTS: Significant improvements in depression (HAMD) and anxiety (HAM-A) were observed at the six-week endpoint, with remission and response rates greater than 50%. There was a positive correlation between endpoint HAM-A scores and HAM-D scores, r = 0.80, (p < 0.01). Improvements were realized at low serum lithium concentrations (0.49 ± 0.20 mEq/L). LIMITATIONS: Lack of placebo control and small sample size warrants validation in larger randomized studies. CONCLUSIONS: Taken in the context of prior evidence, lithium may have an important role in treating comorbid anxiety in bipolar disorder, both as adjunct and monotherapy. Lower doses of lithium may provide equivalent efficacy and enhance tolerability and compliance.

Pain assessment in horses using automatic facial expression recognition through deep learning-based modeling.

The aim of this study was to develop and evaluate a machine vision algorithm to assess the pain level in horses, using an automatic computational classifier based on the Horse Grimace Scale (HGS) and trained by machine learning method. The use of the Horse Grimace Scale is dependent on a human observer, who most of the time does not have availability to evaluate the animal for long periods and must also be well trained in order to apply the evaluation system correctly. In addition, even with adequate training, the presence of an unknown person near an animal in pain can result in behavioral changes, making the evaluation more complex. As a possible solution, the automatic video-imaging system will be able to monitor pain responses in horses more accurately and in real-time, and thus allow an earlier diagnosis and more efficient treatment for the affected animals. This study is based on assessment of facial expressions of 7 horses that underwent castration, collected through a video system positioned on the top of the feeder station, capturing images at 4 distinct timepoints daily for two days before and four days after surgical castration. A labeling process was applied to build a pain facial image database and machine learning methods were used to train the computational pain classifier. The machine vision algorithm was developed through the training of a Convolutional Neural Network (CNN) that resulted in an overall accuracy of 75.8% while classifying pain on three levels: not present, moderately present, and obviously present. While classifying between two categories (pain not present and pain present) the overall accuracy reached 88.3%. Although there are some improvements to be made in order to use the system in a daily routine, the model appears promising and capable of measuring pain on images of horses automatically through facial expressions, collected from video images.

Simultaneous determination of cadmium, lead and copper in chocolate samples by square wave anodic stripping voltammetry.

In this work, an effective and simple method is proposed for the simultaneous determination of cadmium, lead and copper in chocolate samples by Square Wave Anodic Stripping Voltammetry (SWASV). An ultrasonic bath was used for the extraction of cadmium, lead and copper from fourteen chocolate samples using HNO3 solution (7 mol L-1). The electrochemical system consisted of a cell with three electrodes and HCl solution (10 mmol L-1) as the supporting electrolyte. An efficient extraction of the metals (~100%) was attained after 1 h of ultrasonic pre-treatment. Quantitative analysis was carried out by the standard addition method. Good linearity, precision and accuracy were obtained in the range of concentrations examined. The accuracy was evaluated by means of a reference sample of spiked skim milk powder (BCR 151) to prove the reliability of the method. Detection limits (LOD) of 0.089, 0.059 and 0.018 µg g-1 were found for cadmium, copper and lead, respectively, in the chocolate samples. Concentrations in chocolate samples were 4.30-138 µg g-1 for Cu and 0.83-27.9 µg g-1 for Pb, with no significant Cd. The simultaneous determination brings advantages to other methods already reported for chocolate analysis and the samples preparation proposed avoids the traditional sample mineralization step. These characteristics show this new method is especially attractive for case studies and routine analysis.

Multi-Sensor Wearable Health Device Framework for Real-Time Monitoring of Elderly Patients Using a Mobile Application and High-Resolution Parameter Estimation.

Automatized scalable healthcare support solutions allow real-time 24/7 health monitoring of patients, prioritizing medical treatment according to health conditions, reducing medical appointments in clinics and hospitals, and enabling easy exchange of information among healthcare professionals. With recent health safety guidelines due to the COVID-19 pandemic, protecting the elderly has become imperative. However, state-of-the-art health wearable device platforms present limitations in hardware, parameter estimation algorithms, and software architecture. This paper proposes a complete framework for health systems composed of multi-sensor wearable health devices (MWHD), high-resolution parameter estimation, and real-time monitoring applications. The framework is appropriate for real-time monitoring of elderly patients' health without physical contact with healthcare professionals, maintaining safety standards. The hardware includes sensors for monitoring steps, pulse oximetry, heart rate (HR), and temperature using low-power wireless communication. In terms of parameter estimation, the embedded circuit uses high-resolution signal processing algorithms that result in an improved measure of the HR. The proposed high-resolution signal processing-based approach outperforms state-of-the-art HR estimation measurements using the photoplethysmography (PPG) sensor.

Downregulated genes by silencing MYC pathway identified with RNA-SEQ analysis as potential prognostic biomarkers in gastric adenocarcinoma.

MYC overexpression is a common phenomenon in gastric carcinogenesis. In this study, we identified genes differentially expressed with a downregulated profile in gastric cancer (GC) cell lines with silenced MYC. The TTLL12, CDKN3, CDC16, PTPRA, MZT2B, UBE2T genes were validated using qRT-PCR, western blot and immunohistochemistry in tissues of 213 patients with diffuse and intestinal GC. We identified high levels of TTLL12, MZT2B, CDC16, UBE2T, associated with early and advanced stages, lymph nodes, distant metastases and risk factors such as H. pylori. Our results show that in the diffuse GC the overexpression of CDC16 and UBE2T indicate markers of poor prognosis higher than TTLL12. That is, patients with overexpression of these two genes live less than patients with overexpression of TTLL12. In the intestinal GC, patients who overexpressed CDC16 had a significantly lower survival rate than patients who overexpressed MZT2B and UBE2T, indicating in our data a worse prognostic value of CDC16 compared to the other two genes. PTPRA and CDKN3 proved to be important for assessing tumor progression in the early and advanced stages. In summary, in this study, we identified diagnostic and prognostic biomarkers of GC under the control of MYC, related to the cell cycle and the neoplastic process.

Methodological quality and risk of bias of systematic reviews about loading time of multiple dental implants in totally or partially edentulous patients: An umbrella systematic review.

BACKGROUND: There are several systematic reviews of multiple implant loading techniques, but results are conflicting. AIM: To perform an umbrella review on methodological quality of systematic reviews about techniques for loading multiple dental implants. MATERIAL AND METHODS: MEDLINE (PubMed) and Scopus were searched up to December 31, 2019. Unpublished literature was searched through OpenGray and references of included articles were manually verified. Eligibility criteria were: articles had to (1) be about multiple dental implants; (2) mention the moment of loading; (3) be a systematic review. Two independent reviewers participated in the entire process. Qualitative description of included studies as well as methodological quality measurement and risk of bias through AMSTAR and ROBIS were performed. RESULTS: 21 reviews were included. Thirteen stated that there was a similarity between loading techniques, two did not affirm which one was more appropriate and six mentioned that conventional technique was better. Eight papers were classified as high risk of bias, twelve as low and one as uncertain risk. CONCLUSION: When evaluating only studies with a low risk of bias, there are no significant differences in implant loading time.

Anonymous Real-Time Analytics Monitoring Solution for Decision Making Supported by Sentiment Analysis.

Currently, social networks present information of great relevance to various government agencies and different types of companies, which need knowledge insights for their business strategies. From this point of view, an important technique for data analysis is to create and maintain an environment for collecting data and transforming them into intelligence information to enable analysts to observe the evolution of a given topic, elaborate the analysis hypothesis, identify botnets, and generate data to aid in the decision-making process. Focusing on collecting, analyzing, and supporting decision-making, this paper proposes an architecture designed to monitor and perform anonymous real-time searches in tweets to generate information allowing sentiment analysis on a given subject. Therefore, a technological structure and its implementation are defined, followed by processes for data collection and analysis. The results obtained indicate that the proposed solution provides a high capacity to collect, process, search, analyze, and view a large number of tweets in several languages, in real-time, with sentiment analysis capabilities, at a low cost of implementation and operation.

Direct analysis of ascorbic acid in food beverage samples by flow injection analysis using reduced graphene oxide sensor.

In this paper, a simple, sensitive and precise electroanalytical method was developed using flow injection analysis (FIA) with amperometric detection and reduced graphene oxide sensor for ascorbic acid determination in samples of multivitamin beverages, milk, fermented milk, and milk chocolate. The advantages of this sensor include a potential displacement of 450 mV and a 2-fold peak current increase for electrochemical oxidation of ascorbic acid, which resulted in a highly sensitive method. No interference of sample matrix was observed, avoiding solvent extraction procedures (samples were only diluted). The FIA allowed a high analytical frequency, approximately 96 injections per hour, together with adequate detection limit of 4.7 µmol L-1. Good precision (RSD < 7%) and accuracy (recoveries between 91 and 108%) evidenced the robustness of the method. The method was compared with ultra-fast liquid chromatography (UFLC) obtaining statistically similar results (95% confidence level). The ascorbic acid content in samples varied from 0.065 to 2.53 mmol L-1.

Data Security and Trading Framework for Smart Grids in Neighborhood Area Networks.

Due to the drastic increase of electricity prosumers, i.e., energy consumers that are also producers, smart grids have become a key solution for electricity infrastructure. In smart grids, one of the most crucial requirements is the privacy of the final users. The vast majority of the literature addresses the privacy issue by providing ways of hiding user's electricity consumption. However, open issues in the literature related to the privacy of the electricity producers still remain. In this paper, we propose a framework that preserves the secrecy of prosumers' identities and provides protection against the traffic analysis attack in a competitive market for energy trade in a Neighborhood Area Network (NAN). In addition, the amount of bidders and of successful bids are hidden from malicious attackers by our framework. Due to the need for small data throughput for the bidders, the communication links of our framework are based on a proprietary communication system. Still, in terms of data security, we adopt the Advanced Encryption Standard (AES) 128 bit with Exclusive-OR (XOR) keys due to their reduced computational complexity, allowing fast processing. Our framework outperforms the state-of-the-art solutions in terms of privacy protection and trading flexibility in a prosumer-to-prosumer design.

Chromium Levels in Tobacco, Filter and Ash of Illicit Brands Cigarettes Marketed in Brazil.

Smoking is a public health problem and an important source of exposure to toxic metals. This work describes an efficient analytical method comparable to the ones based on atomic emission techniques for the determination of chromium in different constituent parts of cigarette samples (tobacco, filters and ashes) using electrothermal vaporization-atomic absorption spectrometry. The method was evaluated using 12 samples, and the results showed recovery values between 83 and 107%. The accuracy was also evaluated using a reference sample of tomato leaves (NIST SRM 1573a), which proved the efficiency of the method. The limits of detection of the developed method were 20.4, 75.8 and 80.7 ng g-1 for tobacco, filter and cigarette ash samples, respectively. The average chromium values found for the analyzed samples were in the range of 0.96 to 3.85 and from 0.32 to 0.80 µg/cigarette for tobacco and ashes, respectively. For most pre-burn and post-burn filter samples, the values of chromium concentration remained below limits of detection. The developed method presented adequate results about precision and accuracy, demonstrating its applicability in the determination of chromium in cigarette samples.

Effective Mass of Quasiparticles in Armchair Graphene Nanoribbons.

Armchair graphene nanoribbons (AGNRs) may present intrinsic semiconducting bandgaps, being of potential interest in developing new organic-based optoelectronic devices. The induction of a bandgap in AGNRs results from quantum confinement effects, which reduce charge mobility. In this sense, quasiparticles' effective mass becomes relevant for the understanding of charge transport in these systems. In the present work, we theoretically investigate the drift of different quasiparticle species in AGNRs employing a 2D generalization of the Su-Schrieffer-Heeger Hamiltonian. Remarkably, our findings reveal that the effective mass strongly depends on the nanoribbon width and its value can reach 60 times the mass of one electron for narrow lattices. Such underlying property for quasiparticles, within the framework of gap tuning engineering in AGNRs, impact the design of their electronic devices.

Tensor-Based Subspace Tracking for Time-Delay Estimation in GNSS Multi-Antenna Receivers.

Although Global Navigation Satellite Systems (GNSS) receivers currently achieve high accuracy when processing their geographic location under line of sight (LOS), multipath interference and noise degrades the accuracy considerably. In order to mitigate multipath interference, receivers based on multiple antennas became the focus of research and technological development. In this context, tensor-based approaches based on Parallel Factor Analysis (PARAFAC) models have been proposed in the literature, providing optimum performance. State-of-the-art techniques for antenna array based GNSS receivers compute singular value decomposition (SVD) for each new sample, implying into a high computational complexity, being, therefore, prohibitive for real-time applications. Therefore, in order to reduce the computational complexity of the parameter estimates, subspace tracking algorithms are essential. In this work, we propose a tensor-based subspace tracking framework to reduce the overall computational complexity of the highly accurate tensor-based time-delay estimation process.

Evaluation of the autonomic nervous system by analysis of heart rate variability in the preterm infants.

BACKGROUND: Premature infants may present with damage to the autonomic nervous system (ANS), which may be related to poorer neurological development. Among the techniques used to evaluate the ANS, heart rate variability (HRV) emerged as a simple, non-invasive, and easy to apply tool. The aim of the present study was to analyze and compare HRV in preterm infants at different times of hospitalization in order to verify the possible environmental relationships or clinical evolution with HRV. METHODS: A longitudinal, prospective, and descriptive study with non-probabilistic sampling composed of 25 collections of preterm infants of HRV at two moments: moment I (within 15 days of birth) and moment II (after 45 days post-birth). The Polar V800 heart rate monitor was used with the Polar H10 cardiac transducer to collect HRV, which was collected in the supine position for 15 min. The HRV data were analyzed by the linear method in frequency domain and time domain and by the nonlinear method using Kubios HRV analysis software, version 3.0.2. RESULTS: There was an increase in HRV values at moment II, these being statistically significant in the SD1, ApEn, and SampEn. Data related to increased sympathetic nervous system activity, parasympathetic nervous system activity, and increased index complexity. CONCLUSIONS: The data demonstrate an increase in HRV values in premature infants at moment II, demonstrating a possible development in the maturation of the ANS during hospitalization. TRIAL REGISTRATION: RBR-3x7gz8 retrospectively registered.