157 research outputs found
Report on the colloquium, L’aristocratie odryse : signes et lieux du pouvoir en Thrace (Ve-IIIe siècles avant J.-C.), Paris, 12-13 juin 2015
No full textSideris Athanasios. Report on the colloquium, L’aristocratie odryse : signes et lieux du pouvoir en Thrace (Ve-IIIe siècles avant J.-C.), Paris, 12-13 juin 2015. In: Dialogues d'histoire ancienne, vol. 41, n°2, 2015. pp. 202-206
Report on the colloquium, L’aristocratie odryse : signes et lieux du pouvoir en Thrace (Ve-IIIe siècles avant J.-C.), Paris, 12-13 juin 2015
No full textSideris Athanasios. Report on the colloquium, L’aristocratie odryse : signes et lieux du pouvoir en Thrace (Ve-IIIe siècles avant J.-C.), Paris, 12-13 juin 2015. In: Dialogues d'histoire ancienne, vol. 41, n°2, 2015. pp. 202-206
A Lydian Silver Amphora with Zoomorphic Handles
No full textThe paper presents an unpublished silver amphora with zoomorphic handles shaped as winged horses. It
explores the entire class of similar metal amphorae with zoomorphic handles (including isolated handles)
from the Achaemenid domain and its neighbouring areas, as well as their clay parallels, and their icono‑
graphic occurrences. Some possible Greek bronze parallels are also discussed. The vase weight is understood
as following the Achaemenid monetary standard. Finally, the vase is attributed to a Lydian workshop and
dated ca. 500 BC
Recommended from our members
A Visual Tracking Study and A Proposal of Modifications
No full textOn-line visual tracking of a specified target in motion throughout frames of video clips faces challenges in robust identification of the target in the current frame based on the past frames. Three approaches for tracking the target image patch are described and compared. These approaches utilize particle filtering and principal component analysis (PCA) to identify the most likely location of the target in the current frame and a low dimensional subspace representation of the patches of images to be kept as the templates in the dictionary for the identification. By using a combination of methods and compare the result of each, a new model based is proposed. The goal is to achieve a more robust and accurate tracking of a target throughout the video and continue updating the identification templates to adapt the target changes, such as apparences in lighting, angle, scale and occlusions. The challenges in tracking are to introduction of the "right" templates into the identification templates in the dictionary and identify the most accurate particle image patch while tracking the target with the right tracking patch scaling. The first approach considered and on which the structure of the visual tracker is based is the "Incremental Learning for Robust Visual Tracking" by D. Ross et al., which is a computationally fast tracker that utilizes a method of low dimensional subspace for the identification template dictionary and incremental PCA for its tracking. The tracker has a simple rule in accepting the patches of images to be in the identification template dictionary after the image patch has gone through a singular value decomposition (SVD), where it eliminates singular values are smaller than of the sum of squared sinuglar values and the corresponding bases are also eliminated. This elimination scheme has very limited robustness in tracking, therefore, more selective processes in accepting identification templates in the dictionary are explored and introduced on top of the existing method in comparison and to address the challenges in on-line video tracking. The second approach is the "Least Soft-Threshold Squares Tracking" proposed by D. Wang et al. solves the least soft-threshold squares distance problem to identify the distances of the particles to the templates in the dictionary, which greatly improves the tracking accuracy. This method is also computationally cheap in comparison to the first approach, and its accuracy is also better than the first approach, but it would sometimes fail to track in some applications. Finally, the third approach reviewed is the "Robust Visual Tracking and Vehicle Classification via Sparse Representation" by X. Mei et al. is to weight each particles when selecting the most likely target patch so the best patch has a highest weighted probability which ensures it being selected and introduced to the template dictionary. This approach performs well in comparison to the first and the second approaches in tracking accuracy and robustness, but this approach is extremely computationally expensive. Three new components are proposed in an effort to mitigate some of the limitations that the three approaches exhibit. One such component is to simply reject the image patches that exhibit too great of difference to the current template dictionary, which resulted in improved tracking robustness. This method is computationally cheap and easy to implement. Another component introduced is a second set of dictionary that is composed of admitted image patches, which is used for tracking when the image patches appears to be too dissimilar to the dictionary with low dimensional representation. It is expected that with more well defined and stronger features, it forces the tracking to identify the target. Finally, the third component introduced is the to prevent shrinkage of the target boundary box by weighting the particles drawn with the ratio of area change so that more weight is placed on particles with less arial change. This increases the likelihood of recovering the target again if tracking loses the target, and instead of shrinking the boundary box, the tracking is biased to staying with the image patch of the same size. The resulting performance of the proposed tracking scheme has not been noticeably improved, part of the reason is because the metrics available to identify a noisy image patch from the good image patches are not always indicative of the noisy-good image patch divide
Managing atrial fibrillation in the very elderly patient: challenges and solutions
No full textNikolaos Karamichalakis, Konstantinos P Letsas, Konstantinos Vlachos, Stamatis Georgopoulos, Athanasios Bakalakos, Michael Efremidis, Antonios Sideris Laboratory of Cardiac Electrophysiology, "Evangelismos" General Hospital of Athens, Athens, Greece Abstract: Atrial fibrillation (AF) is the most common arrhythmia affecting elderly patients. Management and treatment of AF in this rapidly growing population of older patients involve a comprehensive assessment that includes comorbidities, functional, and social status. The cornerstone in therapy of AF is thromboembolic protection. Anticoagulation therapy has evolved, using conventional or newer medications. Percutaneous left atrial appendage closure is a new invasive procedure evolving as an alternative to systematic anticoagulation therapy. Rate or rhythm control leads to relief in symptoms, fewer hospitalizations, and an improvement in quality of life. Invasive methods, such as catheter ablation, are the new frontier of treatment in maintaining an even sinus rhythm in this particular population. Keywords: elderly, atrial fibrillation, anticoagulation, drugs, catheter ablatio
Recommended from our members
A Sequential Linear Quadratic Approach for Constrained Nonlinear Optimal Control with Adaptive Time Discretization and Application to Higher Elevation Mars Landing Problem
No full textA sequential quadratic programming method is proposed for solving nonlinear optimal control problems subject to general path constraints including mixed state-control and state only constraints. The proposed algorithm further develops on the approach proposed in [1] with objective to eliminate the use of a high number of time intervals for arriving at an optimal solution. This is done by introducing an adaptive time discretization to allow formation of a desirable control profile without utilizing a lot of intervals. The use of fewer time intervals reduces the computation time considerably. This algorithm is further used in this thesis to solve a trajectory planning problem for higher elevation Mars landing
Recommended from our members
Optimal Structure Discovery of Deep Neural Networks During Training and its Application to Reinforcement Learning
No full textThis thesis presents a series of novel and theoretically grounded approaches to neural network optimization through simultaneous training and pruning to discover optimal structures of deep neural networks. In view of the ever-increasing computational demands of neural networks, we aim to address the critical challenge of balancing computational cost and performance. We propose an algorithm for simultaneous learning and pruning capable of identifying and eliminating irrelevant units/filters of neural networks during the early stages of training. In this manner, we can significantly reduce the computational load during both training and inference. The algorithm is shown to be easily applicable to existing neural network architectures and machine learning applications such as deep reinforcement learning. Our approach is based on variational inference principles using Gaussian scale mixture priors on the neural network weights. The variational posterior distribution of Bernoulli random variables multiplying the units/filters is learned, similarly to adaptive dropout. Pruning occurs when a variational parameter converges to 0 rendering the corresponding structure permanently inactive, thus saving computations during subsequent iterations of training and prediction. The careful construction of a novel “flattening” hyper-prior distribution over the prior parameters is crucial for their optimal selection and addresses challenges such as premature pruning and robustness with respect to different initializations and network sizes. We extend our approach to layer-level pruning, which reduces the sequential computation in neural networks. Here, the variational posterior distributions of scalar Bernoulli random variables multiplying whole layer weight matrices are learned, similarly to adaptive layer-wise dropout. Unit/filter and layer pruning are combined carefully into a single algorithm that optimally balances network performance, pruning levels, and computational as well as memory complexity. This approach introduces a novel cost function that integrates prediction accuracy and network pruning objectives in a complexity-aware manner, using only three user-defined parameters that are easy to interpret and tune. We prove that the solutions of the resulting optimization problem found via stochastic gradient descent have scale/Bernoulli parameters at either 0 or 1; hence describe deterministic and optimally regularized networks. For all methods, we carefully analyze the ODE systems underlying the stochastic optimization algorithms and establish domains of attraction around zero for the dynamics of the network parameters. These results provide theoretical support for safely pruning units/filters and/or layers during training and lead to practical pruning conditions. The proposed methods are extensively evaluated on benchmark data sets (MNIST, CIFAR-10/100 and ImageNet) and popular neural network architectures (LeNet, VGG16, and several ResNets), consistently demonstrating improved pruning ratios and competitive test accuracy compared to state-of-the-art techniques. We demonstrate that the combined approach improves upon layer-only or unit-only pruning and favorably competes with existing combined unit/filter and layer pruning algorithms that require pre-trained networks. We integrate our pruning method with common deep reinforcement learning algorithms such as Proximal Policy Optimization, the Twin Delayed Deep Deterministic Gradient and Soft Actor-Critic algorithm. While such algorithms can be overly sensitive to the choice of the network architecture, we aim to discover suitable network architectures during the agent's learning in the environment and hence eliminate the need for the expensive manual tuning of the network architectures. The integration of our method requires only minor modifications to the network architecture such as introducing skip-connections and the random variables multiplying different structures as well as the addition of the needed regularization terms to the existing objective functions. Simulations on typical benchmark problems demonstrate the effectiveness of our method in finding such networks and in reducing the computational load of the control policy networks and value-function networks during the learning process
Recommended from our members
Efficient Reinforcement Learning with Bayesian Optimization
No full textA probabilistic reinforcement learning algorithm is presented for finding control policies in continuous state and action spaces without a prior knowledge of the dynamics. The objective of this algorithm is to learn from minimal amount of interaction with the environment in order to maximize a notion of reward, i.e. a numerical measure of the quality of the resulting state trajectories. Experience from the interactions are used to construct a set of probabilistic Gaussian process (GP) models that predict the resulting state trajectories and the reward from executing a policy on the system. These predictions are used with a technique known as Bayesian optimization to search for policies that promise higher rewards. As more experience is gathered, predictions are made with more confidence and the search for better policies relies less on new interactions with the environment.The computational demand of a GP makes it eventually impractical to use as the number of observations from interacting with the environment increase. Moreover, using a single GP to model different regions that may exhibit disparate behaviors can produce unsatisfactory representations and predictions. One way of mitigating these issues is by partitioning the observation points into different regions each represented by a local GP. With the sequential arrival of the observation points from new experiences, it is necessary to have an adaptive clustering method that can partition the data into an appropriate number of regions. This led to the development of EM+ algorithm presented in the second part of this work, which is an extension to the Expectation Maximization (EM) for the Gaussian mixture models, that assumes no prior knowledge of the number of components.Lastly, an application of the EM+ algorithm to filtering problems is presented. We propose a filtering algorithm that combines the advantages of the well-known particle filter and the mixture of Gaussian filter, while avoiding their issues
Recommended from our members
Neural Network Closure Modeling of Longitudinal Combustion Instability in a Liquid-propellant Rocket Engine
No full textIn this work, neural network (NN)-based models are generated to replace flamelet tables for sub-grid modeling in large-eddy simulations of a single-injector liquid-propellant rocket engine. The NN training process presents an extraordinary challenge. The multi-dimensional combustion instability problem involves multi-scale lengths and characteristic times in an unsteady flow problem with nonlinear acoustics, addressing both transient and dynamic-equilibrium behaviors, superimposed on a turbulent reacting flow with very narrow, moving flame regions. Accurate interpolation between the points of the training data becomes vital. Computationally efficient and accurate flamelet models for a turbulent combustor are needed for useful large eddy simulations (LES). Promise is offered through the use of deep learning NN. Here, a well-studied configuration through prior LES and experiment is used but now with NN providing the sub-grid model for the flamelets. It is of special interest to extend the use of NN modeling for unsteady behavior of mean pressure and velocity fields.
Two different approaches for NN design are proposed, based on the source of training data used in the NN development. In the first approach, data in the flamelet libraries are used as a source for training NNs to replace those flamelet libraries in the CFD simulation of a turbulent diffusion flame with unsteady pressure.
The models are validated on those libraries and verified by being implemented into CFD simulations. Both transient and dynamic equilibrium oscillatory conditions are considered by varying initial conditions of the simulations. Two different geometrical configurations are also tested. The NN-based simulations are favorably compared with their table-based counterparts.
The NNs proposed based on the second approach are trained based on data processed from a few CFD simulations of a single-injector liquid-propellant rocket engine with different dynamical configurations to reproduce the information stored in a flamelet table. The training set is also enriched by data from the physical characteristics and considerations of the combustion model. Flame temperature is used as an extra input for other flame variables to improve the NN-based model accuracy and physical consistency. The trained NNs are first tested offline on the flamelet table. These physics-aware NN-based closure models are successfully implemented into CFD simulations and verified by being tested on various dynamical configurations. The results from those tests are in good agreement with their counterpart table-based CFD simulations
A Polynomial time Algorithm for Checking the Robust Stability of a Polytope of Polynomials
No full text- …
