Filters








30 Hits in 6.6 sec

Softmax Dissection: Towards Understanding Intra- and Inter-class Objective for Embedding Learning [article]

Lanqing He, Zhongdao Wang, Yali Li, Shengjin Wang
2020 arXiv   pre-print
The softmax loss and its variants are widely used as objectives for embedding learning, especially in applications like face recognition.  ...  Training with regular-scale data, experiments in face verification show D-Softmax is favorably comparable to existing losses such as SphereFace and ArcFace.  ...  and Philbin 2015; or adding angular margin to the softmax loss (Liu et al. 2017; Wang et al. 2018; Deng et al. 2018) .  ... 
arXiv:1908.01281v2 fatcat:ckijkiglqzbe5pxagcu74yefei

Softmax Dissection: Towards Understanding Intra- and Inter-Class Objective for Embedding Learning

Lanqing He, Zhongdao Wang, Yali Li, Shengjin Wang
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
The softmax loss and its variants are widely used as objectives for embedding learning applications like face recognition.  ...  In this paper, we propose to dissect Softmax into independent intra- and inter-class objective (D-Softmax) with a clear understanding.  ...  and Philbin 2015; or adding angular margin to the softmax loss (Liu et al. 2017; Wang et al. 2018; Deng et al. 2018) .  ... 
doi:10.1609/aaai.v34i07.6729 fatcat:h5mi5l5qbfgzxmu324b7fdqhwy

Face Recognition and Tracking Framework for Human–Robot Interaction

Aly Khalifa, Ahmed A. Abdelrahman, Dominykas Strazdas, Jan Hintz, Thorsten Hempel, Ayoub Al-Hamadi
2022 Applied Sciences  
To maintain the accuracy, a single-shot multi-level face localization in the wild (RetinaFace) is utilized for face detection, and additive angular margin loss (ArcFace) is employed for recognition.  ...  In addition, we implemented our system as a modular ROS package that makes it straightforward for integration in different real-world HRI systems.  ...  The ArcFace or Additive Angular Margin Loss [4] is one of the most potent loss functions designed for deep face recognition [41] [42] [43] .  ... 
doi:10.3390/app12115568 fatcat:2piyjsheczhs5i5hxqvkxu3inq

What comprises a good talking-head video generation?: A Survey and Benchmark [article]

Lele Chen, Guofeng Cui, Ziyi Kou, Haitian Zheng, Chenliang Xu
2020 arXiv   pre-print
In this work, we present a carefully-designed benchmark for evaluating talking-head video generation with standardized dataset pre-processing strategies.  ...  As for evaluation, we either propose new metrics or select the most appropriate ones to evaluate results in what we consider as desired properties for a good talking-head video, namely, identity preserving  ...  ArcFace is trained with arc loss, which consists of two parts-Softmax loss and additive angular margin. Then they compute the cosine distance of two vectors for measuring identity mismatch.  ... 
arXiv:2005.03201v1 fatcat:4e7lrttoyzfgfdgcvurroxbsuq

#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning [article]

Haoran Tang, Rein Houthooft, Davis Foote, Adam Stooke, Xi Chen, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel
2017 arXiv   pre-print
Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs  ...  States are mapped to hash codes, which allows to count their occurrences with a hash table.  ...  We gratefully acknowledge the support of the NSF through grant IIS-1619362 and of the ARC through a Laureate Fellowship (FL110100281) and through the ARC Centre of Excellence for Mathematical and Statistical  ... 
arXiv:1611.04717v3 fatcat:pt5t26wxhrcc7kl3ieo3jcvkgu

Learning to Unknot [article]

Sergei Gukov, James Halverson, Fabian Ruehle, Piotr Sułkowski
2020 arXiv   pre-print
Trust region policy optimization (TRPO) performs consistently well for a wide range of crossing numbers and thoroughly outperformed other RL algorithms and random walkers.  ...  Perhaps surprisingly, we find that accuracy increases with the length of the braid word, and that the networks learn a direct correlation between the confidence of their predictions and the degree of the  ...  In other words, while problems in class P are the ones for which an answer can be found in polynomial time, problems in class NP are the ones for which checking the answer can be done in polynomial time  ... 
arXiv:2010.16263v1 fatcat:4pgj4gq2hjgkjb23dmfwiv3wvi

Point2Cyl: Reverse Engineering 3D Objects from Point Clouds to Extrusion Cylinders [article]

Mikaela Angelina Uy, Yen-yu Chang, Minhyuk Sung, Purvi Goel, Joseph Lambourne, Tolga Birdal, Leonidas Guibas
2022 arXiv   pre-print
Precisely, our approach first predicts per-point segmentation, base/barrel labels and normals, then estimates for the underlying extrusion parameters in differentiable and closed-form formulations.  ...  A.7.3 Ablation on Additional Loss Functions We further ablate on adding additional loss functions.  ...  We use the angular error (E.A.) for the extrusion axis, the distance/L2 loss for the extrusion center (E.C.), L1 for the extrusion scale, and L sketch for g DP . A.5.  ... 
arXiv:2112.09329v2 fatcat:3urrlpk6tbeuximtqlswrmbgae

A Contrastive Framework for Learning Sentence Representations from Pairwise and Triple-wise Perspective in Angular Space

Yuhao Zhang, Hongji Zhu, Yongliang Wang, Nan Xu, Xiaobo Li, Binqiang Zhao
2022 Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)   unpublished
Recently, several contrastive learning methods have been proposed for learning sentence representations and have shown promising results.  ...  So in this paper, we propose a new method ArcCSE, with training objectives designed to enhance the pairwise discriminative power and model the entailment relation of triplet sentences.  ...  For pairwise sentence relation modeling, we propose Additive Angular Margin Contrastive Loss (ArcCon Loss), which enhances the pairwise discriminative power by maximizing the decision margin in the angular  ... 
doi:10.18653/v1/2022.acl-long.336 fatcat:eoyqvrxfojhk5c3vilpihh4b2u

Efficient and High Performing Biometrics: Towards Enabling Recognition in Embedded Domains

Fadi Boutros
2022
Towards enhancing face recognition accuracy, this thesis presents a novel margin-penalty softmax loss, ElasticFace, that relaxes the restriction of having a single fixed penalty margin.  ...  The growing need for reliable and accurate recognition solutions along with the recent innovations in deep learning methodologies has reshaped the research landscape of biometric recognition.  ...  The randomized margin penalty can be easily integrated into any of the angular margin-based softmax losses, which we demonstrate on two SOTA margin-based softmax losses.  ... 
doi:10.26083/tuprints-00021571 fatcat:5jedbyc2fnea5gtlakwfhan67m

Surround-View Cameras based Holistic Visual Perception for Automated Driving [article]

Varun Ravi Kumar
2022 arXiv   pre-print
We concentrate on the following issues in order to address them: 1) Developing near-field perception algorithms with high performance and low computational complexity for various visual perception tasks  ...  The dynamics changed from a primitive organism waiting for the food to come into contact for eating food being sought after by visual sensors.  ...  We construct the loss for the additional backward sequence similar to the forward sequence.  ... 
arXiv:2206.05542v1 fatcat:cdpn6afpvvf7hnsvry7cqbjq3u

Quantum computing enhanced machine learning for physico-chemical applications [article]

Manas Sajjan, Junxu Li, Raja Selvarajan, Shree Hari Sureshbabu, Sumit Suresh Kale, Rishabh Gupta, Sabre Kais
2021 arXiv   pre-print
Machine learning (ML) has emerged into formidable force for identifying hidden but pertinent patterns within a given data set with the objective of subsequent generation of automated predictive behavior  ...  Not only the classical variants of ML , even those trainable on near-term quantum hardwares have been developed with promising outcomes.  ...  The output layer being a classifier uses a softmax.  ... 
arXiv:2111.00851v1 fatcat:i2caiglszvbufbyfmf3cwkcduu

CARS 2021: Computer Assisted Radiology and Surgery Proceedings of the 35th International Congress and Exhibition Munich, Germany, June 21–25, 2021

2021 International Journal of Computer Assisted Radiology and Surgery  
The University Rovira i Virgili also supports this work with project 2019PFR-B2-61. References  ...  By combining one of the CE and the softmax losses with the original box loss, we have two loss functions for YOLO-v3 training.  ...  Compare with the MSE and softmax losses, the CE loss has larger output when misclassification happened, which is better for optimization of the detection part in YOLO-v3.  ... 
doi:10.1007/s11548-021-02375-4 pmid:34085172 fatcat:6d564hsv2fbybkhw4wvc7uuxcy

Reinforcement Learning [chapter]

Ke-Lin Du, M. N. S. Swamy
2013 Neural Networks and Statistical Learning  
For example, one approach is based on the theory of partially-observable MDPs (POMDPs).  ...  This probability distribution can then be used as a new state signal for the original problem.  ...  , grasping, and retrieving the box.  ... 
doi:10.1007/978-1-4471-5571-3_18 fatcat:l3xi3uyhtff35gai7qvvvedwbq

Reinforcement Learning [chapter]

Andrew G. Barto
1997 Neural Systems for Control  
For example, one approach is based on the theory of partially-observable MDPs (POMDPs).  ...  This probability distribution can then be used as a new state signal for the original problem.  ...  , grasping, and retrieving the box.  ... 
doi:10.1016/b978-012526430-3/50003-9 fatcat:nwyl3dsnrbc75csknm6u34nyty

Engineering, Technology & Applied Science Research (ETASR), Vol. 11, No. 2, pp. 6845-7068 [article]

Various
2021 Zenodo  
Engineering, Technology & Applied Science Research (ETASR) is an international bimonthly wide scope, peer-reviewed open access journal for the publication of original articles concerned with diverse aspects  ...  The table of contents for this issue can be found at: https://etasr.com/index.php/ETASR/issue/view/68. For more info visit the journal's site at: www.etasr.com.  ...  ACKNOWLEDGMENT The authors are grateful to the Institute for Research and Development, Suan Sunandha Rajabhat University and the Faculty of Science at Ubon Ratchathani University, for supporting this research  ... 
doi:10.5281/zenodo.4720665 fatcat:mk2prflstjaa3bhkjenwy22s6u
« Previous Showing results 1 — 15 out of 30 results