Abstract
In order to achieve a virtual certification process and robust designs for turbomachinery, the uncertainty bounds for computational fluid dynamics have to be known. The formulation of turbulence closure models implies a major source of the overall uncertainty of Reynold-averaged Navier Stokes simulations. We discuss the common practice of applying a physics constrained eigenspace perturbation of the Reynolds stress tensor in order to account for the model form uncertainty of turbulence models. Since the basic methodology often leads to generous uncertainty estimates, we extend a recent approach of adding a machine learning strategy. The application of a data-driven method is motivated by striving for the detection of flow regions, which are prone to suffer from a lack of turbulence model prediction accuracy. In this way any user input related to choosing the degree of uncertainty is supposed to become obsolete. This work especially investigates an approach, which tries to determine an a priori estimation of prediction confidence, when there is no accurate data available to judge the prediction. The flow around the NACA 4412 airfoil at near-stall conditions serves to demonstrate the successful application of the data-driven eigenspace perturbation framework. We especially highlight the objectives and limitations of the underlying methodology finally.
Contour-Aware Equipotential Learning for Semantic Segmentation
Abstract
With increasing demands for high-quality semantic segmentation in the industry, hard-distinguishing semantic boundaries have posed a significant threat to existing solutions. Inspired by real-life experience, i.e., combining varied observations contributes to higher visual recognition confidence, we present the equipotential learning (EPL) method. This novel module transfers the predicted/ground-truth semantic labels to a self-defined potential domain to learn and infer decision boundaries along customized directions. The conversion to the potential domain is implemented via a lightweight differentiable anisotropic convolution without incurring any parameter overhead. Besides, the designed two loss functions, the point loss and the equipotential line loss implement anisotropic field regression and category-level contour learning, respectively, enhancing prediction consistencies in the inter/intra-class boundary areas. More importantly, EPL is agnostic to network architectures, and thus it can be plugged into most existing segmentation models. This paper is the first attempt to address the boundary segmentation problem with field regression and contour learning. Meaningful performance improvements on Pascal Voc 2012 and Cityscapes demonstrate that the proposed EPL module can benefit the off-the-shelf fully convolutional network models when recognizing semantic boundary areas. Besides, intensive comparisons and analysis show the favorable merits of EPL for distinguishing semantically-similar and irregular-shaped categories.
VFLens: Co-design the Modeling Process for Efficient Vertical Federated Learning via Visualization
Authors: Yun Tian, He Wang, Laixin Xie, Xiaojuan Ma, Quan Li
Abstract
As a decentralized training approach, federated learning enables multiple organizations to jointly train a model without exposing their private data. This work investigates vertical federated learning (VFL) to address scenarios where collaborating organizations have the same set of users but with different features, and only one party holds the labels. While VFL shows good performance, practitioners often face uncertainty when preparing non-transparent, internal/external features and samples for the VFL training phase. Moreover, to balance the prediction accuracy and the resource consumption of model inference, practitioners require to know which subset of prediction instances is genuinely needed to invoke the VFL model for inference. To this end, we co-design the VFL modeling process by proposing an interactive real-time visualization system, VFLens, to help practitioners with feature engineering, sample selection, and inference. A usage scenario, a quantitative experiment, and expert feedback suggest that VFLens helps practitioners boost VFL efficiency at a lower cost with sufficient confidence.
Abstract
We present Self-Tuning Tube-based Model Predictive Control (STT-MPC), an adaptive robust control algorithm for uncertain linear systems with additive disturbances based on the least-squares estimator and polytopic tubes. Our algorithm leverages concentration results to bound the system uncertainty set with prescribed confidence, and guarantees robust constraint satisfaction for this set, along with recursive feasibility and input-to-state stability. Persistence of excitation is ensured without compromising the algorithm's asymptotic performance or increasing its computational complexity. We demonstrate the performance of our algorithm using numerical experiments.
Square-root regret bounds for continuous-time episodic Markov decision processes
Authors: Xuefeng Gao, Xun Yu Zhou
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC)
Abstract
We study reinforcement learning for continuous-time Markov decision processes (MDPs) in the finite-horizon episodic setting. We present a learning algorithm based on the methods of value iteration and upper confidence bound. We derive an upper bound on the worst-case expected regret for the proposed algorithm, and establish a worst-case lower bound, both bounds are of the order of square-root on the number of episodes. Finally, we conduct simulation experiments to illustrate the performance of our algorithm.
Sequential Brick Assembly with Efficient Constraint Satisfaction
Authors: Seokjun Ahn, Jungtaek Kim, Minsu Cho, Jaesik Park
Abstract
We address the problem of generating a sequence of LEGO brick assembly with high-fidelity structures, satisfying physical constraints between bricks. The assembly problem is challenging since the number of possible structures increases exponentially with the number of available bricks, complicating the physical constraints to satisfy across bricks. To tackle this problem, our method performs a brick structure assessment to predict the next brick position and its confidence by employing a U-shaped sparse 3D convolutional network. The convolution filter efficiently validates physical constraints in a parallelizable and scalable manner, allowing to process of different brick types. To generate a novel structure, we devise a sampling strategy to determine the next brick position by considering attachable positions under physical constraints. Instead of using handcrafted brick assembly datasets, our model is trained with a large number of 3D objects that allow to create a new high-fidelity structure. We demonstrate that our method successfully generates diverse brick structures while handling two different brick types and outperforms existing methods based on Bayesian optimization, graph generative model, and reinforcement learning, all of which are limited to a single brick type.
Keyword: scaling
Long-Tailed Class Incremental Learning
Authors: Xialei Liu, Yu-Song Hu, Xu-Sheng Cao, Andrew D. Bagdanov, Ke Li, Ming-Ming Cheng
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
In class incremental learning (CIL) a model must learn new classes in a sequential manner without forgetting old ones. However, conventional CIL methods consider a balanced distribution for each new task, which ignores the prevalence of long-tailed distributions in the real world. In this work we propose two long-tailed CIL scenarios, which we term ordered and shuffled LT-CIL. Ordered LT-CIL considers the scenario where we learn from head classes collected with more samples than tail classes which have few. Shuffled LT-CIL, on the other hand, assumes a completely random long-tailed distribution for each task. We systematically evaluate existing methods in both LT-CIL scenarios and demonstrate very different behaviors compared to conventional CIL scenarios. Additionally, we propose a two-stage learning baseline with a learnable weight scaling layer for reducing the bias caused by long-tailed distribution in LT-CIL and which in turn also improves the performance of conventional CIL due to the limited exemplars. Our results demonstrate the superior performance (up to 6.44 points in average incremental accuracy) of our approach on CIFAR-100 and ImageNet-Subset. The code is available at https://github.com/xialeiliu/Long-Tailed-CIL
A Dual Realization of Chua's Chaotic Oscillator Using a Current-Controlled Nonlinear Resistor
Abstract
A dual realization of Chuas chaotic oscillator is proposed using current-controlled nonlinear resistors, one linear resistor, one capacitor and two inductors. Two problems are solved. First, unit rescaling is necessary when transforming the standard chaotic equations into circuit equations to ensure that the current units are milliamperes. In addition, the connection and parameters of two current-controlled nonlinear resistors are set to build the required volt-ampere characteristics. The inductor currents show the classical characteristics of being sensitive to the circuit parameters and initial conditions. In addition, experimental verification is performed to demonstrate the feasibility of the circuit. Chuas dual circuit exhibits rich dynamic chaotic features and might bring new applications due to chaotic currents.
Scaling Laws for a Multi-Agent Reinforcement Learning Model
Abstract
The recent observation of neural power-law scaling relations has made a significant impact in the field of deep learning. A substantial amount of attention has been dedicated as a consequence to the description of scaling laws, although mostly for supervised learning and only to a reduced extent for reinforcement learning frameworks. In this paper we present an extensive study of performance scaling for a cornerstone reinforcement learning algorithm, AlphaZero. On the basis of a relationship between Elo rating, playing strength and power-law scaling, we train AlphaZero agents on the games Connect Four and Pentago and analyze their performance. We find that player strength scales as a power law in neural network parameter count when not bottlenecked by available compute, and as a power of compute when training optimally sized agents. We observe nearly identical scaling exponents for both games. Combining the two observed scaling laws we obtain a power law relating optimal size to compute similar to the ones observed for language models. We find that the predicted scaling of optimal neural network size fits our data for both games. This scaling law implies that previously published state-of-the-art game-playing models are significantly smaller than their optimal size, given the respective compute budgets. We also show that large AlphaZero models are more sample efficient, performing better than smaller models with the same amount of training data.
MSRL: Distributed Reinforcement Learning with Dataflow Fragments
Authors: Huanzhou Zhu, Bo Zhao, Gang Chen, Weifeng Chen, Yijie Chen, Liang Shi, Peter Pietzuch, Lei Chen
Abstract
Reinforcement learning~(RL) trains many agents, which is resource-intensive and must scale to large GPU clusters. Different RL training algorithms offer different opportunities for distributing and parallelising the computation. Yet, current distributed RL systems tie the definition of RL algorithms to their distributed execution: they hard-code particular distribution strategies and only accelerate specific parts of the computation (e.g. policy network updates) on GPU workers. Fundamentally, current systems lack abstractions that decouple RL algorithms from their execution. We describe MindSpore Reinforcement Learning (MSRL), a distributed RL training system that supports distribution policies that govern how RL training computation is parallelised and distributed on cluster resources, without requiring changes to the algorithm implementation. MSRL introduces the new abstraction of a fragmented dataflow graph, which maps Python functions from an RL algorithm's training loop to parallel computational fragments. Fragments are executed on different devices by translating them to low-level dataflow representations, e.g. computational graphs as supported by deep learning engines, CUDA implementations or multi-threaded CPU processes. We show that MSRL subsumes the distribution strategies of existing systems, while scaling RL training to 64 GPUs.
Towards Exascale for Wind Energy Simulations
Authors: Misun Min, Michael Brazell, Ananias Tomboulides, Matthew Churchfield, Paul Fischer, Michael Sprague
Subjects: Computational Engineering, Finance, and Science (cs.CE); Distributed, Parallel, and Cluster Computing (cs.DC); Performance (cs.PF)
Abstract
We examine large-eddy-simulation modeling approaches and computational performance of two open-source computational fluid dynamics codes for the simulation of atmospheric boundary layer (ABL) flows that are of direct relevance to wind energy production. The first is NekRS, a high-order, unstructured-grid, spectral element code. The second, AMR-Wind, is a block-structured, second-order finite-volume code with adaptive-mesh-refinement capabilities. The objective of this study is to co-develop these codes in order to improve model fidelity and performance for each. These features will be critical for running ABL-based applications such as wind farm analysis on advanced computing architectures. To this end, we investigate the performance of NekRS and AMR-Wind on the Oak Ridge Leadership Facility supercomputers Summit, using 4 to 800 nodes (24 to 4,800 NVIDIA V100 GPUs), and Crusher, the testbed for the Frontier exascale system using 18 to 384 Graphics Compute Dies on AMD MI250X GPUs. We compare strong- and weak-scaling capabilities, linear solver performance, and time to solution. We also identify leading inhibitors to parallel scaling.
Keyword: calibration
Calibrating Sequence likelihood Improves Conditional Language Generation
Authors: Yao Zhao, Misha Khalman, Rishabh Joshi, Shashi Narayan, Mohammad Saleh, Peter J. Liu
Abstract
Conditional language models are predominantly trained with maximum likelihood estimation (MLE), giving probability mass to sparsely observed target sequences. While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. This has been empirically observed in beam search decoding as output quality degrading with large beam sizes, and decoding strategies benefiting from heuristics such as length normalization and repetition-blocking. In this work, we introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model's latent space. With SLiC, decoding heuristics become unnecessary and decoding candidates' quality significantly improves regardless of the decoding method. Furthermore, SLiC shows no sign of diminishing returns with model scale, and presents alternative ways to improve quality with limited training and inference budgets. With SLiC, we exceed or match SOTA results on a wide range of generation tasks spanning abstractive summarization, question generation, abstractive question answering and data-to-text generation, even with modest-sized models.
Learnable Distribution Calibration for Few-Shot Class-Incremental Learning
Authors: Binghao Liu, Boyu Yang, Lingxi Xie, Ren Wang, Qi Tian, Qixiang Ye
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
Few-shot class-incremental learning (FSCIL) faces challenges of memorizing old class distributions and estimating new class distributions given few training samples. In this study, we propose a learnable distribution calibration (LDC) approach, with the aim to systematically solve these two challenges using a unified framework. LDC is built upon a parameterized calibration unit (PCU), which initializes biased distributions for all classes based on classifier vectors (memory-free) and a single covariance matrix. The covariance matrix is shared by all classes, so that the memory costs are fixed. During base training, PCU is endowed with the ability to calibrate biased distributions by recurrently updating sampled features under the supervision of real distributions. During incremental learning, PCU recovers distributions for old classes to avoid forgetting', as well as estimating distributions and augmenting samples for new classes to alleviateover-fitting' caused by the biased distributions of few-shot samples. LDC is theoretically plausible by formatting a variational inference procedure. It improves FSCIL's flexibility as the training procedure requires no class similarity priori. Experiments on CUB200, CIFAR100, and mini-ImageNet datasets show that LDC outperforms the state-of-the-arts by 4.64%, 1.98%, and 3.97%, respectively. LDC's effectiveness is also validated on few-shot learning scenarios.
Unsupervised Vision and Vision-motion Calibration Strategies for PointGoal Navigation in Indoor Environment
Authors: Yijun Cao, Xianshi Zhang, Fuya Luo, Yongjie Li
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
PointGoal navigation in indoor environment is a fundamental task for personal robots to navigate to a specified point. Recent studies solved this PointGoal navigation task with near-perfect success rate in photo-realistically simulated environments, under the assumptions with noiseless actuation and most importantly, perfect localization with GPS and compass sensors. However, accurate GPS signal can not be obtained in real indoor environment. To improve the pointgoal navigation accuracy in real indoor, we proposed novel vision and vision-motion calibration strategies to train visual and motion path integration in unsupervised manner. Sepecifically, visual calibration computes the relative pose of the agent from the re-projection error of two adjacent frames, and then replaces the accurate GPS signal with the path integration. This pseudo position is also used to calibrate self-motion integration which assists agent to update their internal perception of location and helps improve the success rate of navigation. The training and inference process only use RGB, depth, collision as well as self-action information. The experiments show that the proposed system achieves satisfactory results and outperforms the partially supervised learning algorithms on the popular Gibson dataset.
Fast and Robust Video-Based Exercise Classification via Body Pose Tracking and Scalable Multivariate Time Series Classifiers
Authors: Ashish Singh, Antonio Bevilacqua, Thach Le Nguyen, Feiyan Hu, Kevin McGuinness, Martin OReilly, Darragh Whelan, Brian Caulfield, Georgiana Ifrim
Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Abstract
Technological advancements have spurred the usage of machine learning based applications in sports science. Physiotherapists, sports coaches and athletes actively look to incorporate the latest technologies in order to further improve performance and avoid injuries. While wearable sensors are very popular, their use is hindered by constraints on battery power and sensor calibration, especially for use cases which require multiple sensors to be placed on the body. Hence, there is renewed interest in video-based data capture and analysis for sports science. In this paper, we present the application of classifying S\&C exercises using video. We focus on the popular Military Press exercise, where the execution is captured with a video-camera using a mobile device, such as a mobile phone, and the goal is to classify the execution into different types. Since video recordings need a lot of storage and computation, this use case requires data reduction, while preserving the classification accuracy and enabling fast prediction. To this end, we propose an approach named BodyMTS to turn video into time series by employing body pose tracking, followed by training and prediction using multivariate time series classifiers. We analyze the accuracy and robustness of BodyMTS and show that it is robust to different types of noise caused by either video quality or pose estimation factors. We compare BodyMTS to state-of-the-art deep learning methods which classify human activity directly from videos and show that BodyMTS achieves similar accuracy, but with reduced running time and model engineering effort. Finally, we discuss some of the practical aspects of employing BodyMTS in this application in terms of accuracy and robustness under reduced data quality and size. We show that BodyMTS achieves an average accuracy of 87\%, which is significantly higher than the accuracy of human domain experts.
Leveraging unsupervised data and domain adaptation for deep regression in low-cost sensor calibration
Abstract
Air quality monitoring is becoming an essential task with rising awareness about air quality. Low cost air quality sensors are easy to deploy but are not as reliable as the costly and bulky reference monitors. The low quality sensors can be calibrated against the reference monitors with the help of deep learning. In this paper, we translate the task of sensor calibration into a semi-supervised domain adaptation problem and propose a novel solution for the same. The problem is challenging because it is a regression problem with covariate shift and label gap. We use histogram loss instead of mean squared or mean absolute error, which is commonly used for regression, and find it useful against covariate shift. To handle the label gap, we propose weighting of samples for adversarial entropy optimization. In experimental evaluations, the proposed scheme outperforms many competitive baselines, which are based on semi-supervised and supervised domain adaptation, in terms of R2 score and mean absolute error. Ablation studies show the relevance of each proposed component in the entire scheme.
Extrinsic calibration for highly accurate trajectories reconstruction
Authors: Maxime Vaidis, William Dubois, Alexandre Guénette, Johann Laconte, Vladimír Kubelka, François Pomerleau
Abstract
In the context of robotics, accurate ground-truth positioning is the cornerstone for the development of mapping and localization algorithms. In outdoor environments and over long distances, total stations provide accurate and precise measurements, that are unaffected by the usual factors that deteriorate the accuracy of Global Navigation Satellite System (GNSS). While a single robotic total station can track the position of a target in three Degrees Of Freedom (DOF), three robotic total stations and three targets are necessary to yield the full six DOF pose reference. Since it is crucial to express the position of targets in a common coordinate frame, we present a novel extrinsic calibration method of multiple robotic total stations with field deployment in mind. The proposed method does not require the manual collection of ground control points during the system setup, nor does it require tedious synchronous measurement on each robotic total station. Based on extensive experimental work, we compare our approach to the classical extrinsic calibration methods used in geomatics for surveying and demonstrate that our approach brings substantial time savings during the deployment. Tested on more than 30 km of trajectories, our new method increases the precision of the extrinsic calibration by 25 % compared to the best state-of-the-art method, which is the one taking manually static ground control points.
Keyword: out of distribution detection
There is no result
Keyword: out-of-distribution detection
There is no result
Keyword: expected calibration error
There is no result
Keyword: overconfident
There is no result
Keyword: overconfidence
There is no result
Keyword: confidence
Evaluation of physics constrained data-driven methods for turbulence model uncertainty quantification
Contour-Aware Equipotential Learning for Semantic Segmentation
VFLens: Co-design the Modeling Process for Efficient Vertical Federated Learning via Visualization
Self-Tuning Tube-based Model Predictive Control
Square-root regret bounds for continuous-time episodic Markov decision processes
Sequential Brick Assembly with Efficient Constraint Satisfaction
Keyword: scaling
Long-Tailed Class Incremental Learning
A Dual Realization of Chua's Chaotic Oscillator Using a Current-Controlled Nonlinear Resistor
Scaling Laws for a Multi-Agent Reinforcement Learning Model
MSRL: Distributed Reinforcement Learning with Dataflow Fragments
Towards Exascale for Wind Energy Simulations
Keyword: calibration
Calibrating Sequence likelihood Improves Conditional Language Generation
Learnable Distribution Calibration for Few-Shot Class-Incremental Learning
forgetting', as well as estimating distributions and augmenting samples for new classes to alleviate
over-fitting' caused by the biased distributions of few-shot samples. LDC is theoretically plausible by formatting a variational inference procedure. It improves FSCIL's flexibility as the training procedure requires no class similarity priori. Experiments on CUB200, CIFAR100, and mini-ImageNet datasets show that LDC outperforms the state-of-the-arts by 4.64%, 1.98%, and 3.97%, respectively. LDC's effectiveness is also validated on few-shot learning scenarios.Unsupervised Vision and Vision-motion Calibration Strategies for PointGoal Navigation in Indoor Environment
Fast and Robust Video-Based Exercise Classification via Body Pose Tracking and Scalable Multivariate Time Series Classifiers
Leveraging unsupervised data and domain adaptation for deep regression in low-cost sensor calibration
Extrinsic calibration for highly accurate trajectories reconstruction