scholarly journals 3D cephalometric landmark detection by multiple stage deep reinforcement learning

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Sung Ho Kang ◽  
Kiwan Jeon ◽  
Sang-Hoon Kang ◽  
Sang-Hwy Lee
Keyword(s):  
Reinforcement Learning ◽  
Three Dimensional ◽  
Detection Accuracy ◽  
Sequential Decision ◽  
Landmark Detection ◽  
Boundary Estimation ◽  
Multi Stage ◽  
Sequential Decision Process ◽  
Gradient Based ◽  
3D Cephalometry

AbstractThe lengthy time needed for manual landmarking has delayed the widespread adoption of three-dimensional (3D) cephalometry. We here propose an automatic 3D cephalometric annotation system based on multi-stage deep reinforcement learning (DRL) and volume-rendered imaging. This system considers geometrical characteristics of landmarks and simulates the sequential decision process underlying human professional landmarking patterns. It consists mainly of constructing an appropriate two-dimensional cutaway or 3D model view, then implementing single-stage DRL with gradient-based boundary estimation or multi-stage DRL to dictate the 3D coordinates of target landmarks. This system clearly shows sufficient detection accuracy and stability for direct clinical applications, with a low level of detection error and low inter-individual variation (1.96 ± 0.78 mm). Our system, moreover, requires no additional steps of segmentation and 3D mesh-object construction for landmark detection. We believe these system features will enable fast-track cephalometric analysis and planning and expect it to achieve greater accuracy as larger CT datasets become available for training and testing.

An Efficient Multiple Description Coding for Multi-View Video Based on the Correlation of Spatial Polyphase Transformed Subsequences

2019 ◽  
Vol 63 (5) ◽  
pp. 50401-1-50401-7 ◽  
Author(s):  
Jing Chen ◽  
Jie Liao ◽  
Huanqiang Zeng ◽  
Canhui Cai ◽  
Kai-Kuang Ma
Keyword(s):  
Video Transmission ◽  
Video Sequence ◽  
Three Dimensional ◽  
The Other ◽  
Multiple Description Coding ◽  
Multiple Description ◽  
Prediction Mode ◽  
Gradient Based ◽  
Directional Interpolation ◽  
Description Coding

Abstract For a robust three-dimensional video transmission through error prone channels, an efficient multiple description coding for multi-view video based on the correlation of spatial polyphase transformed subsequences (CSPT_MDC_MVC) is proposed in this article. The input multi-view video sequence is first separated into four subsequences by spatial polyphase transform and then grouped into two descriptions. With the correlation of macroblocks in corresponding subsequence positions, these subsequences should not be coded in completely the same way. In each description, one subsequence is directly coded by the Joint Multi-view Video Coding (JMVC) encoder and the other subsequence is classified into four sets. According to the classification, the indirectly coding subsequence selectively employed the prediction mode and the prediction vector of the counter directly coding subsequence, which reduces the bitrate consumption and the coding complexity of multiple description coding for multi-view video. On the decoder side, the gradient-based directional interpolation is employed to improve the side reconstructed quality. The effectiveness and robustness of the proposed algorithm is verified by experiments in the JMVC coding platform.

scholarly journals Selective network discovery via deep reinforcement learning on embedded spaces

2021 ◽  
Vol 6 (1) ◽  
Author(s):  
Peter Morales ◽  
Rajmonda Sulo Caceres ◽  
Tina Eliassi-Rad
Keyword(s):  
Reinforcement Learning ◽  
Learning Algorithm ◽  
Sequential Decision ◽  
Network Discovery ◽  
Learning Tasks ◽  
Partially Observed ◽  
Decision Making Problem ◽  
Resource Collection ◽  
Improved Performance ◽  
Discovery Algorithms

AbstractComplex networks are often either too large for full exploration, partially accessible, or partially observed. Downstream learning tasks on these incomplete networks can produce low quality results. In addition, reducing the incompleteness of the network can be costly and nontrivial. As a result, network discovery algorithms optimized for specific downstream learning tasks given resource collection constraints are of great interest. In this paper, we formulate the task-specific network discovery problem as a sequential decision-making problem. Our downstream task is selective harvesting, the optimal collection of vertices with a particular attribute. We propose a framework, called network actor critic (NAC), which learns a policy and notion of future reward in an offline setting via a deep reinforcement learning algorithm. The NAC paradigm utilizes a task-specific network embedding to reduce the state space complexity. A detailed comparative analysis of popular network embeddings is presented with respect to their role in supporting offline planning. Furthermore, a quantitative study is presented on various synthetic and real benchmarks using NAC and several baselines. We show that offline models of reward and network discovery policies lead to significantly improved performance when compared to competitive online discovery algorithms. Finally, we outline learning regimes where planning is critical in addressing sparse and changing reward signals.

scholarly journals Optimal Policies for Quantum Markov Decision Processes

2021 ◽  
Author(s):  
Ming-Sheng Ying ◽  
Yuan Feng ◽  
Sheng-Gang Ying
Keyword(s):  
Decision Making ◽  
Reinforcement Learning ◽  
Quantum Systems ◽  
Sequential Decision Making ◽  
Mathematical Framework ◽  
Sequential Decision ◽  
Learning Techniques ◽  
Optimal Policies ◽  
Markov Decision ◽  
Programming Algorithms

AbstractMarkov decision process (MDP) offers a general framework for modelling sequential decision making where outcomes are random. In particular, it serves as a mathematical framework for reinforcement learning. This paper introduces an extension of MDP, namely quantum MDP (qMDP), that can serve as a mathematical model of decision making about quantum systems. We develop dynamic programming algorithms for policy evaluation and finding optimal policies for qMDPs in the case of finite-horizon. The results obtained in this paper provide some useful mathematical tools for reinforcement learning techniques applied to the quantum world.

Blade-Row Interactions in an LP Turbine at Design and Strong Off-Design Operation

2014 ◽  
Author(s):  
Martin Lipfert ◽  
Jan Habermann ◽  
Martin G. Rose ◽  
Stephan Staudacher ◽  
Yavuz Guendogdu
Keyword(s):  
Experimental Data ◽  
Flow Field ◽  
Three Dimensional ◽  
Suction Side ◽  
Test Facility ◽  
Time Resolved ◽  
Joint Project ◽  
Multi Stage ◽  
Blade Row ◽  
Wake Interactions

In a joint project between the Institute of Aircraft Propulsion Systems (ILA) and MTU Aero Engines a two-stage low pressure turbine is tested at design and strong off-design conditions. The experimental data taken in the altitude test-facility aims to study the effect of positive and negative incidence of the second stator vane. A detailed insight and understanding of the blade row interactions at these regimes is sought. Steady and time-resolved pressure measurements on the airfoil as well as inlet and outlet hot-film traverses at identical Reynolds number are performed for the midspan streamline. The results are compared with unsteady multi-stage CFD predictions. Simulations agree well with the experimental data and allow detailed insights in the time-resolved flow-field. Airfoil pressure field responses are found to increase with positve incidence whereas at negative incidence the magnitude remains unchanged. Different pressure to suction side phasing is observed for the studied regimes. The assessment of unsteady blade forces reveals that changes in unsteady lift are minor compared to changes in axial force components. These increase with increasing positive incidence. The wake-interactions are predominating the blade responses in all regimes. For the positive incidence conditions vane 1 passage vortex fluid is involved in the midspan passage interaction leading to a more distorted three-dimensional flow field.

scholarly journals A Gradient-Based Reinforcement Learning Algorithm for Multiple Cooperative Agents

IEEE Access ◽  
2018 ◽  
Vol 6 ◽  
pp. 70223-70235 ◽  
Author(s):  
Zhen Zhang ◽  
Dongqing Wang ◽  
Dongbin Zhao ◽  
Qiaoni Han ◽  
Tingting Song
Keyword(s):  
Reinforcement Learning ◽  
Learning Algorithm ◽  
Cooperative Agents ◽  
Gradient Based ◽  
Reinforcement Learning Algorithm

Inverse Design of 3D Multi-Stage Transonic Fans at Dual Operating Points

2013 ◽  
Author(s):  
James H. Page ◽  
Paul Hield ◽  
Paul G. Tucker
Keyword(s):  
Design Method ◽  
Pressure Ratio ◽  
Three Dimensional ◽  
Inverse Design ◽  
Flow Angle ◽  
Trade Offs ◽  
Multi Stage ◽  
Full Speed ◽  
Computational Fluid Dynamics Cfd ◽  
Operating Points

Semi-inverse design is the automatic re-cambering of an aerofoil, during a computational fluid dynamics (CFD) calculation, in order to achieve a target lift distribution while maintaining thickness, hence “semi-inverse”. In this design method, the streamwise distribution of curvature is replaced by a stream-wise distribution of lift. The authors have developed an inverse design code based on the method of Hield (2008) which can rapidly design three-dimensional fan blades in a multi-stage environment. The algorithm uses an inner loop to design to radially varying target lift distributions, an outer loop to achieve radial distributions of stage pressure ratio and exit flow angle, and a choked nozzle to set design mass flow. The code is easily wrapped around any CFD solver. In this paper, we describe a novel algorithm for designing simultaneously for specified performance at full speed and peak efficiency at part speed, without trade-offs between the targets at each of the two operating points. We also introduce a novel adaptive target lift distribution which automatically develops discontinuous changes of calculated magnitude, based on the passage shock, eliminating erroneous lift demands in the shock vicinity and maintaining a smooth aerofoil.

A High Fidelity Quality Assessment of High Pressure Turbine Blades Using Surface Curvature and Gradient-Based Adjoint

2018 ◽  
Author(s):  
Akhil Mulloth ◽  
Gabriel Banks ◽  
Giulio Zamboni ◽  
Simon Bather
Keyword(s):  
Quality Assessment ◽  
Turbine Blades ◽  
Three Dimensional ◽  
Surface Curvature ◽  
Assessment Process ◽  
Screening Process ◽  
Profile Tolerance ◽  
Gradient Based ◽  
Parts Manufacturing ◽  
The Impact

Gas turbine performance is highly dependent on the quality of the manufactured parts. Manufacturing variations in the parts can significantly alter the performance, especially efficiency and thus SFC. The legacy process is to accept variations within predefined profile tolerance limits and a few other qualitative parameters, mostly at a few, key two-dimensional aerofoil sections. With the widespread use of White light scans and other similar three-dimensional scans, this has improved to include the three-dimensional profile. The future however may lie with performance based quality assessment of manufactured parts, combined with quantitative surface quality assessment to implement an intelligent screening process for the parts. The adjoint method, typically used for shape optimization is adapted to provide a prediction of the impact on performance due to manufacturing variations. The work presented outlines a three stage quality assessment process for manufactured parts, involving three-dimensional profile tolerance based screening, followed by a surface curvature based screening and finally an Adjoint based performance prediction.

Sign in / Sign up

Export Citation Format

Share Document