Filters








113,112 Hits in 4.3 sec

Detached Error Feedback for Distributed SGD with Random Sparsification [article]

An Xu, Heng Huang
2022 arXiv   pre-print
Then all the workers communication to get the average gradient g t = 1 K K k=1 g k,t , where K is the total number of workers, and update the model via θ t+1 = θ t − ηg t , ( 2 ) where η is the learning  ...  One is the "query" y t which is updated via SGD, and the other is the "key" x t (x 0 = y 0 ) which is updated via x t+1 = (1 − δ)x t + δy t . ( 30 ) The success of momentum contrast is explained as a "  ...  ., y t+1 = y t − η t ∇f (y t ; ξ t ) = y t − η t g t . (95) Let {x t } be the SGD-IA solution path with x 0 = y 0 , where IA denotes momentum iterative averaging, i.e., where 0 < 1 − δ < 1 is the momentum  ... 
arXiv:2004.05298v3 fatcat:mrb5anlugfhdpopmtz3rkg3ddy

Sensor Networks Routing via Bayesian Exploration

Shuang Hao, Ting Wang
2006 Local Computer Networks (LCN), Proceedings of the IEEE Conference on  
To this end, we deploy a Bayesian method to offer good balance between exploitation and exploration.  ...  It estimates the benefit of exploration by value of information therefore avoids the error-prone process of parameter tuning which usually requires human intervention.  ...  The Bayesian exploration framework smoothly integrates model extraction with prior knowledge and optimal exploration.  ... 
doi:10.1109/lcn.2006.322207 dblp:conf/lcn/HaoW06 fatcat:ow2yj7ksv5eenpyoyn33y3vf74

Hierarchical Exploration for Accelerating Contextual Bandits [article]

Yisong Yue, Sue Ann Hong (Carnegie Mellon University), Carlos Guestrin
2012 arXiv   pre-print
Contextual bandit learning is an increasingly popular approach to optimizing recommender systems via user feedback, but can be slow to converge in practice due to the need for exploring a large feature  ...  Intuitively, user preferences can be reasonably embedded in a coarse low-dimensional feature space that can be explored efficiently, requiring exploration in the high-dimensional space only as necessary  ...  It is difficult to optimize (13) directly, so we approximate it using a smooth formulation, 4 argmin U ∈span(U0):U 2 F ro =K w∈W w 2 , (14) where we now constrain U via U 2 F ro = K.  ... 
arXiv:1206.6454v1 fatcat:h2zoninebfhptnfttrpwxqkg5a

Selective Zero-Shot Classification with Augmented Attributes [article]

Jie Song, Chengchao Shen, Jie Lei, An-Xiang Zeng, Kairi Ou, Dacheng Tao, Mingli Song
2018 arXiv   pre-print
Finally, the prediction confidence is measured by both the defined and the residual attributes.  ...  We propose a selective zero-shot classifier based on both the human defined and the automatically discovered residual attributes.  ...  The proposed classifier explores and exploits the residual properties beyond the defined attributes for defining confidence functions.  ... 
arXiv:1807.07437v1 fatcat:5xxgmv33wfbylcrjnfm3nl4nxy

Sustainable intensification of crop residue exploitation for bioenergy: opportunities and challenges

Ioanna Mouratiadou, Tommaso Stella, Thomas Gaiser, Birka Wicke, Claas Nendel, Frank Ewert, Floor van der Hilst
2019 GCB Bioenergy  
We explore opportunities to increase bioenergy potentials from residues while reducing environmental impacts, in line with sustainable intensification.  ...  In order to sustainably intensify crop residue exploitation for bioenergy and reconcile climate change mitigation with other sustainability objectives, such as those on soil and water quality, residue  ...  Using North Rhine-Westphalia (NRW) in Germany as a case study, we first explore stakeholder views on bar riers to sustainable crop residue exploitation.  ... 
doi:10.1111/gcbb.12649 pmid:32025242 pmcid:PMC6988490 fatcat:zry7fskowfdv7a4rez4ztchonq

Reducing self-interference in full duplex transmission by interference exploitation

Mahmoud T. Kabir, Muhammad R. A. Khandaker, Christos Masouros
2017 GLOBECOM 2017 - 2017 IEEE Global Communications Conference  
In this paper, we consider the power minimization problem in a multi-user full-duplex communication system by employing a multi-objective optimization problem via the weighted Tchebycheff method.  ...  We propose to exploit the multiuser interference by using the knowledge of the data symbols and channel state information at the full-duplex base station.  ...  Constructive interference is yet to be explored in the realm of FD communication systems, where FD brings unique opportunities to be explored with respect to existing works on IE.  ... 
doi:10.1109/glocom.2017.8254804 dblp:conf/globecom/KabirKM17 fatcat:ocbjuauktreqxeimhvvjg7iyda

Sample Efficient Reinforcement Learning via Model-Ensemble Exploration and Exploitation [article]

Yao Yao, Li Xiao, Zhicheng An, Wanpeng Zhang, Dijun Luo
2021 arXiv   pre-print
To mitigate these issues, we present MEEE, a model-ensemble method that consists of optimistic exploration and weighted exploitation.  ...  During exploration, unlike prior methods directly selecting the optimal action that maximizes the expected accumulative return, our agent first generates a set of action candidates and then seeks out the  ...  To sum up, using uncertainty-weighted exploitation or optimism-based exploration alone sometimes degrades performance slightly.  ... 
arXiv:2107.01825v1 fatcat:fkzgnwkaxbblxkgn4jgqcfwqni

Wide-Slice Residual Networks for Food Recognition [article]

Niki Martinel, Gian Luca Foresti, Christian Micheloni
2016 arXiv   pre-print
Specifically, inspired by the recent success of residual deep network, we exploit such a learning scheme and introduce a slice convolution block to capture the vertical food layers.  ...  Outputs of the deep residual blocks are combined with the sliced convolution to produce the classification score for specific food categories.  ...  Residual@WISeR shows the performance achieved via the residual learning branch.  ... 
arXiv:1612.06543v1 fatcat:zkyrdg6t2rdqvdr5sl2ogriigm

Conceiving, exploring, and exploiting innovative ideas: from waste cooking oil to diesel

Stella Bezergianni
2013 Journal of Innovation and Entrepreneurship  
This is readily shown in Figure 2 , where the WCO and hydrotreated WCO products are compared via their distillation curve, which in essence shows the size distribution of the contained molecules.  ...  The idea was conceived, explored, and exploited aiming to get commercialized and promote environmental and economic benefits.  ... 
doi:10.1186/2192-5372-2-9 fatcat:hpuujiavofb7xbzdnybctqlhwi

Adaptive Clustering based Dynamic Routing of Wireless Sensor Networks via Generalized Ant Colony Optimization

Zhengmao Ye, Habib Mohamadian
2014 Information Engineering Research Institute procedia  
Each sensor node estimates the residual energy and dynamically calculates probabilities to select an optimal channel to extend the lifespan of WSNs.  ...  Dynamic clustering based routing is proposed to achieve good performance via adaptive algorithms.  ...  The local update and global update will both applied to the ACO problem to achieve a balance between exploration and exploitation.  ... 
doi:10.1016/j.ieri.2014.09.063 fatcat:7vbytvrjrbf4zbktbgymwsizxm

SIRe-Networks: Skip Connections over Interlaced Multi-Task Learning and Residual Connections for Structure Preserving Object Classification [article]

Danilo Avola, Luigi Cinque, Alessio Fagioli, Gian Luca Foresti
2021 arXiv   pre-print
To validate the presented methodology, a simple CNN and various implementations of famous networks are extended via the SIRe strategy and extensively tested on the CIFAR100 dataset; where the SIRe-extended  ...  time, optimizing the network hyper-parameters, or increasing the architecture depth.  ...  multi-task learning strategy; which is further refined via long skip and short residual connections.  ... 
arXiv:2110.02776v1 fatcat:kc7jwglh65ghfn444e6qwf6ez4

Safe Learning MPC with Limited Model Knowledge and Data [article]

Aaron Kandel, Scott J. Moura
2022 arXiv   pre-print
This paper presents a general framework for safe learning and control (LaC) using stochastic MPC and distributionally robust optimization (DRO).  ...  In this paper, we focus on LaC where the controller is applied directly to a system of which it has no direct experience.  ...  The LaC problem space borrows many concepts from historical research on stochastic optimal control, a field which dates back decades to the original linear-quadratic Gaussian problem [1] .  ... 
arXiv:2004.00759v4 fatcat:edo6h3urcjbcrdcn7il6cj6nwy

BioWSN: A Bio-Inspired Method for Optimization of Routing in Wireless Sensor Networks

Ramin Ahmadi, Gholamhossein Ekbatanifard, Peyman Bayat, Jian Li
2022 Mathematical Problems in Engineering  
To overcome this drawback of the conventional GWO, we introduce a balancing factor between the exploration and exploitation phases of the algorithm in addition to a mapping scheme.  ...  In this study, we propose an improved version of grey wolf optimizer (GWO), a nature-inspired metaheuristic optimization algorithm, to perform cluster head selection and routing in WSN while maximizing  ...  When | A → | > 1 search space is being explored and when | A → | < 1 attacking the prey is executed. Improved Grey Wolf Optimizer (IGWO). e two main phases of GWO are exploration and exploitation.  ... 
doi:10.1155/2022/4826388 fatcat:w4fc43ah65dl5l24yx7h2lbnky

An Educational Platform in Structural Mechanics

Tiago A. N. Silva, Maria A. R. Loja
2013 International Journal of Online Engineering (iJOE)  
stress level and distribution along the material and the advantages of using a structural optimization technique in order to minimize the drawback thermal residual stresses effects.  ...  On the order hand, it is also addressed the use of a population based optimization algorithm in order to attain the referred minimum stress level.  ...  assess the application of DE on structural optimization via the minimization of the magnitude of thermal residual stresses in a FGM sandwich panel.  ... 
doi:10.3991/ijoe.v9is8.3319 fatcat:ohxkye6lsrc2vgarfdzqfpwdje

Second-Order Attention Network for Single Image Super-Resolution

Tao Dai, Jianrui Cai, Yongbing Zhang, Shu-Tao Xia, Lei Zhang
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
repeated local-source residual attention groups (LSRAG) to learn increasingly abstract feature representations.  ...  Recently, deep convolutional neural networks (CNNs) have been widely explored in single image super-resolution (SISR) and obtained remarkable performance.  ...  Then the extracted deep feature F DF is upscaled via the upscale module via F ↑ = H ↑ (F DF ), (3) where H ↑ (·) and F ↑ are a upscale module and upscaled feature respectively.  ... 
doi:10.1109/cvpr.2019.01132 dblp:conf/cvpr/DaiCZXZ19 fatcat:tylc6qiuhrbqdju4w5os5vvika
« Previous Showing results 1 — 15 out of 113,112 results