Continuous Set Research Articles

Energy Efficient AP Selection for Cell-Free Massive MIMO Systems: Deep Reinforcement Learning Approach

The problem of access point (AP) to device association in a cell-free massive multiple-input multiple-output (MIMO) system is investigated. Utilizing energy efficiency (EE) as our main metric, we determine the optimal association parameters subject to minimum rate constraints for all devices. We incorporate all existing practical concerns in our formulation, including training errors, pilot contamination, and central processing unit access to only statistical channel state information (CSI). This EE maximization problem is highly non-convex and possibly NP-hard. We propose to solve this challenging problem by model-free deep reinforcement learning (DRL) methods. Due to the very large discrete action space of our posed optimization problem, existing DRL approaches can not be directly applied. Thus, we approximate the large discrete action space with either a continuous set or a smaller discrete set, and modify existing DRL methods accordingly. Our novel approximations offer a framework with tolerable complexity and satisfactory performance that can be readily applied to other challenging optimization problems in wireless communication. Simulation results corroborate the superior performance of the modified DRL methods over conventional approaches.

Systolic inequalities for the number of vertices

Inspired by the classical Riemannian systolic inequality of Gromov, we present a combinatorial analogue providing a lower bound on the number of vertices of a simplicial complex in terms of its edge-path systole. Similarly to the Riemannian case, where the inequality holds under a topological assumption of “essentiality”, our proofs rely on a combinatorial analogue of that assumption. Under a stronger assumption, expressed in terms of cohomology cup-length, we improve our results quantitatively. We also illustrate our methods in the continuous setting, generalizing and improving quantitatively the Minkowski principle of Balacheff and Karam; a corollary of this result is the extension of the Guth–Nakamura cup-length systolic bound from manifolds to complexes.

A Maximum Divergence Approach to Optimal Policy in Deep Reinforcement Learning.

Model-free reinforcement learning algorithms based on entropy regularized have achieved good performance in control tasks. Those algorithms consider using the entropy-regularized term for the policy to learn a stochastic policy. This work provides a new perspective that aims to explicitly learn a representation of intrinsic information in state transition to obtain a multimodal stochastic policy, for dealing with the tradeoff between exploration and exploitation. We study a class of Markov decision processes (MDPs) with divergence maximization, called divergence MDPs. The goal of the divergence MDPs is to find an optimal stochastic policy that maximizes the sum of both the expected discounted total rewards and a divergence term, where the divergence function learns the implicit information of state transition. Thus, it can provide better-off stochastic policies to improve both in robustness and performance in a high-dimension continuous setting. Under this framework, the optimality equations can be obtained, and then a divergence actor-critic algorithm is developed based on the divergence policy iteration method to address large-scale continuous problems. The experimental results, compared to other methods, show that our approach achieved better performance and robustness in the complex environment particularly. The code of DivAC can be found in https://github.com/yzyvl/DivAC.

Total Generalized Variation for Piecewise Constant Functions on Triangular Meshes with Applications in Imaging

We propose a novel discrete concept for the total generalized variation (TGV), which was originally derived to reduce the staircasing effect in classical total variation regularization, in image denoising problems. We describe discrete, second-order TGV for piecewise constant functions on triangular meshes, thus allowing the TGV functional to be applied to more general data structures than pixel images, and in particular in the context of finite element discretizations. Particular attention is given to the description of the kernel of the TGV functional, which, in the continuous setting, consists of linear polynomials. We discuss how to take advantage of this kernel structure using piecewise constant functions on triangular meshes. Numerical experiments include denoising and inpainting problems for images defined on nonstandard grids, including data from a three-dimensional scanner.

The logit dynamic in supermodular games with a continuum of strategies: A deterministic approximation approach

We consider large population supermodular games with pairwise interaction and a continuous strategy set. Our objective is to establish convergence of the logit dynamic in such games to logit equilibria. For this purpose, we apply the deterministic approximation approach, which interprets a deterministic dynamic as an approximation of a stochastic process. We first establish the closeness of this dynamic with a step–wise approximation. We then show that the logit stochastic process is close to the step–wise logit dynamic in a discrete approximation of the original game. Combining the two results, we obtain our deterministic approximation result. We then apply this result to supermodular games. Over finite but sufficiently long time horizons, the logit stochastic process converges to logit equilibria in a discrete approximation of the supermodular game. By the deterministic approximation approach, so does the logit dynamic in the continuum supermodular game.

Acute Management of Hypertriglyceridemia With a Disease-Specific Intravenous Insulin Infusion Order Set.

Hypertriglyceridemia-associated acute pancreatitis is a disease lacking a standardized management approach. The main objective of this study was to evaluate the safety and efficacy of a continuous intravenous insulin infusion order set specifically designed for managing hypertriglyceridemia. This study compared the safety and efficacy of a standardized (postintervention) approach to managing hypertriglyceridemia to a nonstandardized (preintervention) approach. The primary efficacy outcome was the percentage of patients who achieved a triglyceride level less than 500 mg/dL. Additional outcomes included the time to achieving a triglyceride level less than 500 mg/dL and the percent reduction in triglyceride levels. The primary safety outcome was the number of patients who experienced hypoglycemia (glucose less than 70 mg/dL). Twenty patients were included in both the preintervention and postintervention groups. There was a significantly greater reduction in triglyceride levels observed in the postintervention group. The number of patients who achieved a triglyceride level less than 500 mg/dL in the preintervention and postintervention groups were 10 (50%) and 17 (85%), respectively, P = 0.018. Within the postintervention group, the time to achieving a triglyceride level less than 500 mg/dL in those with and without diabetes was 56.8 hours (38.2-64.0) versus 27.6 hours (19.7-45.0), respectively, P = 0.028. Our findings demonstrate that insulin infusions are safe and effective when therapy is standardized and accounts for nursing to patient ratios. Our results provide the medical community with a standardized approach to acutely managing hypertriglyceridemia.

Simpson's Grading Scale for WHO Grade I Meningioma Resection in the Modern Neurosurgical Era: Are We Really Asking the Right Question?

The Simpson grading scale for the classification of the extent of meningioma resection provided a tremendous movement forward in 1957 suggesting increasing the extent of resection improves recurrence rates. However, equal, if not greater, movements forward have been made in the neurosurgical community over the last half a century owing to improvements in neuroimaging capabilities, microsurgical techniques, and radiotherapeutic strategies. Sughrue et al proposed the idea that these advancements have altered what a "recurrence" and "subtotal resection" truly means in modern neurosurgery compared with Simpson's era, and that a mandated use of the Simpson Scale is likely less clinically relevant today. A subsequent period of debate ensued in the literature which sought to re-examine the clinical value of using the Simpson Scale in modern neurosurgery. While a large body of evidence has recently been provided, these data generally continue to support the clinical importance of gross tumor resection as well as the value of adjuvant radiation therapy and the importance of recently updated World Health Organization classifications. However, there remains a negligible interval benefit in performing overly aggressive surgery and heroic maneuvers to remove the last bit of tumor, dura, and/or bone just for the simple act of achieving a lower Simpson score. Ultimately, meningioma surgery may be better contextualized as a continuous set of weighted risk-benefit decisions throughout the entire operation.

On pp-elimination and stability in a continuous setting

We generalize pp-elimination for modules, or more generally, abelian structures, to a continuous logic setting where the abelian structure is equipped with a homomorphism to a compact Hausdorff group. We conclude that the continuous logic theory of such a structure is stable.

Riesz Transform for a Flow Laplacian on Homogeneous Trees

We prove the L^p-boundedness, for p in (1,infty ), of the first order Riesz transform associated to the flow Laplacian on a homogeneous tree with the canonical flow measure. This result was previously proved to hold for p in (1,2] by Hebisch and Steger, but their approach does not extend to p>2 as we make clear by proving a negative endpoint result for p = infty for such operator. We also consider a class of “horizontal Riesz transforms” corresponding to differentiation along horocycles, which inherit all the boundedness properties of the Riesz transform associated to the flow Laplacian, but for which we are also able to prove a weak type (1, 1) bound for the adjoint operators, in the spirit of the work by Gaudry and Sjögren in the continuous setting. The homogeneous tree with the canonical flow measure is a model case of a measure-metric space which is nondoubling, of exponential growth, does not satisfy the Cheeger isoperimetric inequality, and where the Laplacian does not have spectral gap.

Error-Divisible Two-Qubit Gates

We introduce a simple widely applicable formalism for designing ``error-divisible'' two-qubit gates: a quantum gate set where fractional rotations have proportionally reduced error compared to the full entangling gate. In current noisy intermediate-scale quantum (NISQ) algorithms, performance is largely constrained by error proliferation at high circuit depths, of which two-qubit gate error is generally the dominant contribution. Further, in many hardware implementations, arbitrary two-qubit rotations must be composed from multiple two-qubit stock gates, further increasing error. This work introduces a set of criteria, and example wave forms and protocols to satisfy them, using superconducting qubits with tunable couplers for constructing continuous gate sets with significantly reduced leakage and dynamic $ZZ$ errors for small-angle rotations. If implemented at scale, NISQ-algorithm performance could be significantly improved by our error-divisible gate protocols.

Real-Time Drowsiness Detection System for Student Tracking using Machine Learning

Many studies on fatigue detection have been carried out that were focused on experimention over different technologies. Machine vision based driver fatigue detection systems are used to prevent accidents and improve safety on roads. We propose the design of an alerting system for the students that will use real time video of a person to capture the drowsiness level and will signal alert to the student when the student is in that state of fatigue. A device, if enabled with the system, will start the webcam and track the person. An alert will be generated based on the set frame rate when a continuous set of frames are detected as drowsy. The conventional methods cannot capture complex expressions, however the vailability of deep learning models has enabled a substantial research on detection of states of a person in real time. Our system operates in natural lighting conditions and can predict accurately even when the face is covered with glasses, head caps, etc. The system is implemented using YOLOv5 models (You Look Only Once) is an extremely fast and accurate detection model.

Continuous trajectory similarity search with result diversification

With the continued increasing availability of spatio-temporal data from GPS-equipped devices, online map services, and a variety of location-based social media, trajectory similarity search has become a fundamental operation in location-based data analytics. It is of great importance to enable real-time search of trajectories that satisfy users’ personalized requirements. For the purpose, we study the Diversified Continuous Trajectory Similarity Search (DCTSS) problem. The DCTSS problem aims to process a large number of Continuous Location Set (CLS) queries over a stream of trajectory data while taking result diversity of each query into consideration. To answer the DCTSS problem, we develop a Diversity-Aware Trajectory Publish/Subscribe (DAT-PS) framework, which takes both trajectory data streams and CLS queries as input and performs query-trajectory matching between CLS queries in the query collection and trajectories over the trajectory data stream. Our experimental results on two real-life datasets show that our proposed DAT-PS framework is capable of demonstrating substantial superiority regarding both efficiency and scalability compared against baselines.

Measuring forest height from space. Opportunities and limitations observed in natural forests

The goal of the study is to test a methodology that outputs continuous sets of data by integrating forest canopy height satellite measurements in a Random Forest (RF) algorithm. We assess the performance of the two spaceborne LiDAR missions by comparing them to field measurements taken with a mobile LiDAR scanner (MLS), then we test the compatibility of the sensors as an input for an RF model. The Spearman analysis showed a significant correlation between the MLS height and the GEDI height (p = 0.00015), but not for the ICEsat-2 height measurements (p = 0.3). The same analysis proved to be inconclusive for the RF output, where, although the p-value was 0.00025, the R value was only 0.38. Although the method studied in this paper proved to be only fit for estimation purposes, the integration of spaceborne LiDAR sensors in the RF shows high potential for obtaining homogenous, continuous sets of data.

Strain- and stress-based parametric optimization of fiber-reinforced elastomers under finite deformations

In this work, a parametric optimization framework is developed to deal with fiber-reinforced elastomeric matrices. The fiber orientations are in a continuous set to obtain the set of angles that minimizes a functional given by the designer. Although this topic has been extensively studied in recent years, the vast majority of works use either gradient-based or global-based methods to define the composite fiber orientation, with the first being very prone to reach poor local minima and the latter increasing the computational burden exponentially with the number of design variables. Besides that, considering finite strain hyperelasticity, the number of works is substantially reduced. The main goal of this research is to study the efficiency of a local-based derivative-free algorithm in the optimization of problems considering finite deformation and material nonlinearity assumptions in fiber composite design. In order to show the proposed methodology efficiency, two optimization problems are studied: strain energy density minimization and mechanical stress minimization, with the latter further divided into a global and a cluster approach. Therefore, the main novelty of this work is to deal with fiber-reinforced elastomers optimization by means of a direct search algorithm in a constrained nonlinear optimization. The methodology can easily be implemented in existing finite element codes and the results show the pros and cons of both types of optimization.

A Novel Modulation Method Based on Model Prediction Control With Significantly Reduced Switching Loss and Current Zero-Crossing Distortion for Vienna Rectifier

Finite control set model predictive control (FCS-MPC) is widely used in rectifiers due to its fast response. When there are multiple control objectives, the optimal switching sequence is selected by minimizing the cost function with weighting factors. However, the reduction of current ripple caused by output error and switching loss is in conflict to a certain extent for FCS-MPC. In addition, when multiple objectives are controlled, the selection of weighting factor is a serious challenge. Therefore, a modulation method based on continuous control set model predictive control is proposed in this article. Two phases are clamped and one phase is modulated, which not only can significantly reduce the switching loss without increasing the computational burden, but also can control the neutral point voltage without using weighting factors. In addition, when the proposed method is adopted, current zero-crossing distortion can be suppressed over the full range of modulation index. The effectiveness of the proposed method is verified by simulation and experimental results.

Structures and solvent effects on the 1H and 13C NMR chemical shifts of the family of three pro-lithospermate molecules: DFT study

The solvent effects on the structures and nuclear magnetic resonance spectroscopy of a series of three pro-lithospermate molecules are studied. These molecules include methyl pro-lithospermate, ethyl pro-lithospermate, and n-butyl pro-lithospermate, which exhibit enantiomers and diastereoisomers. The trans-oriented enantiomers were observed by Yu et al. [Nat. Prod. Res. 31(21), 2505–2512 (2017)] and are the subject of the present study. The density functional theory was validated using the gauge-including atomic orbital/B3LYP-D3BJ/6–31++G(d,p) and continuous set of gauge transformation (CSGT)/the long-range corrected Tao, Perdew, Staroverov, and Scuseria’s τ-dependent gradient-corrected correlation-exchange functional (LC-TPSS)TPSS/cc-pVTZ methods. The chemical shifts of 1H and 13C were calculated in dimethylsulfoxide (DMSO), methanol, and water. It turns out that solvent effects are weak on the geometrical parameters (bond lengths and bond angles). The harmonic oscillator model of the aromaticity index was used to calculate the aromaticity of the subunit rings. With regard to the available experimental results, the chemical shifts of 1H and 13C are reproduced better by B3LYP-D3BJ/6–31++G(d,p) than by LC-TPSSTPSS/cc-pVTZ in the DMSO medium. The solvent effects are significant on the 13C chemical shift, mostly for protic solvents, and insignificant on 1H. The vicinal spin–spin coupling constants of protons were calculated. The coupling constant of trans-olefinic and ortho-aromatic protons fits well with the experiment.

A wholistic view of continual learning with deep neural networks: Forgotten lessons and the bridge to active and open world learning

Current deep learning methods are regarded as favorable if they empirically perform well on dedicated test sets. This mentality is seamlessly reflected in the resurfacing area of continual learning, where consecutively arriving data is investigated. The core challenge is framed as protecting previously acquired representations from being catastrophically forgotten. However, comparison of individual methods is nevertheless performed in isolation from the real world by monitoring accumulated benchmark test set performance. The closed world assumption remains predominant, i.e. models are evaluated on data that is guaranteed to originate from the same distribution as used for training. This poses a massive challenge as neural networks are well known to provide overconfident false predictions on unknown and corrupted instances. In this work we critically survey the literature and argue that notable lessons from open set recognition, identifying unknown examples outside of the observed set, and the adjacent field of active learning, querying data to maximize the expected performance gain, are frequently overlooked in the deep learning era. Hence, we propose a consolidated view to bridge continual learning, active learning and open set recognition in deep neural networks. Finally, the established synergies are supported empirically, showing joint improvement in alleviating catastrophic forgetting, querying data, selecting task orders, while exhibiting robust open world application.

Obstacle avoidance method for fixed trajectory of a seven-degree-of-freedom manipulator

AbstractThis paper, based on the idea of redundancy angle discretisation, proposes an obstacle avoidance method for the fixed tip pose trajectory of a seven degrees-of-freedom (7-DOF) modular manipulator. First, for the case in which a specific redundancy angle is given, the analytical solutions of the redundant manipulator left 6-DOF subchain are found. Then, through the discretisation of the redundancy angle, the concept of the self-motion space of the tip pose is proposed and is extended to the concept of the self-motion space of the trajectory. Based on this discrete space, a path-planning algorithm is proposed to help select the appropriate redundancy angles to obtain the collision-free solution set of the fixed Cartesian trajectory. However, due to the large fluctuation of the obtained path, a path optimisation method based on the path cost is proposed to smooth the path, and the continuous and collision-free solution set of the manipulator tip’s trajectory is obtained. The method proposed in this paper provides a new thought for the problem of collision-free solution set planning for the Cartesian trajectory of a 7-DOF manipulator and it has great application potential in working environments with high accuracy requirements for the trajectory.

Associations between Early Life Nutrient Intakes and Brain Maturation Show Developmental Dynamics from Infancy to Toddlerhood: A Neuroimaging Observation Study

BackgroundMyelin imaging has increasingly been applied to study the impact of nutrition on brain development in recent years. Although individual dynamics for nutrient intakes and myelin trajectories previously have been investigated across childhood, the longitudinal interaction between both remains unclear in typically developed children. ObjectivesThe objective of this work was to explore the developmental dynamics of nutrient-myelin interactions from infancy to early childhood using myelin imaging as a marker for brain maturation. MethodsBrain neuroimaging (1 scan per child) and dietary nutrient intake data were analyzed for 88 nutrients from 293 children (127 female, 62% White) from a longitudinal cohort study in the United States. A sliding window approach was used to investigate correlations between nutrient intakes and brain myelination over a continuous set of age windows. Image processing techniques (Sobel-filter vertical edge detection) were applied to determine age windows with unique association profiles, providing novel insight into how these relationships change with child age. ResultsWe identified 3 nutrient-myelin windows covering the age range of 1–5 y: window 1 from 6 to 20 mo with 60% positive nutrient correlations, window 2 from 20 to 30 mo with 20% positive correlations, and window 3 from 30 to 60 mo with 37% positive correlations. The windows are aligned with reported myelin and white matter dynamics that change in the first 5 y from fast and steep (window 1) to continued but slower growth (window 3), with window 2 possibly representing the inflection period. ConclusionsTo our knowledge, this is the first study in typically developing children demonstrating the developmental dynamics between early life nutrient intakes and brain maturation in toddlerhood. The knowledge can be applied for identifying targeted and brain-stage–appropriate nutritional interventions for this critical stage of brain development.

Image dataset for benchmarking automated fish detection and classification algorithms

Multiparametric video-cabled marine observatories are becoming strategic to monitor remotely and in real-time the marine ecosystem. Those platforms can achieve continuous, high-frequency and long-lasting image data sets that require automation in order to extract biological time series. The OBSEA, located at 4 km from Vilanova i la Geltrú at 20 m depth, was used to produce coastal fish time series continuously over the 24-h during 2013–2014. The image content of the photos was extracted via tagging, resulting in 69917 fish tags of 30 taxa identified. We also provided a meteorological and oceanographic dataset filtered by a quality control procedure to define real-world conditions affecting image quality. The tagged fish dataset can be of great importance to develop Artificial Intelligence routines for the automated identification and classification of fishes in extensive time-lapse image sets.