4,221 Hits in 10.2 sec

Improved Trainable Calibration Method for Neural Networks on Medical Imaging Classification [article]

Gongbo Liang, Yu Zhang, Xiaoqin Wang, Nathan Jacobs
2020 arXiv   pre-print
We propose a novel calibration approach that maintains the overall classification accuracy while significantly improving model calibration.  ...  Empirically, neural networks are often miscalibrated and overconfident in their predictions.  ...  Trainable Calibration Methods Trainable calibration methods are proposed to integrate model calibration into classification training.  ... 
arXiv:2009.04057v1 fatcat:oved252a6bhbpiofayz6e5cpmm

Fair-Net: A Network Architecture For Reducing Performance Disparity Between Identifiable Sub-Populations [article]

Arghya Datta, S. Joshua Swamidass
2021 arXiv   pre-print
We introduced Fair-Net, a branched multitask neural network architecture that improves both classification accuracy and probability calibration across identifiable sub-populations in class imbalanced datasets  ...  Empirical studies with three real world benchmark datasets demonstrate that Fair-Net improves classification and calibration performance, substantially reducing performance disparity between gender and  ...  Recently, the Cal-Net neural network architecture [14] demonstrated simultaneous improvement in classification and calibration performance on class imbalanced datasets.  ... 
arXiv:2106.00720v2 fatcat:wvrfezjebvatrhz4o6mfrndbze

Learn to Communicate with Neural Calibration: Scalability and Generalization [article]

Yifan Ma, Yifei Shen, Xianghao Yu, Jun Zhang, S.H. Song, Khaled B. Letaief
2021 arXiv   pre-print
Simulation results will show that the proposed neural calibration approach enjoys significantly improved scalability and generalization compared with the existing learning-based methods.  ...  In this paper, we propose a scalable and generalizable neural calibration framework for future wireless system design, where a neural network is adopted to calibrate the input of conventional model-based  ...  Specifically, the backbone of a low-complexity method is retained while neural networks are adopted to calibrate the input and improve system performance.  ... 
arXiv:2110.00272v1 fatcat:crnamzjrkjhsbguckek6s2adc4

Neural Calibration for Scalable Beamforming in FDD Massive MIMO with Implicit Channel Estimation [article]

Yifan Ma, Yifei Shen, Xianghao Yu, Jun Zhang, S.H. Song, Khaled B. Letaief
2021 arXiv   pre-print
Different from the existing fully data-driven approach where all the modules are replaced by deep neural networks (DNNs), a neural calibration method is proposed to improve the scalability of the end-to-end  ...  Simulation results will show the superiority of the proposed neural calibration method over benchmark schemes in terms of both the spectral efficiency and scalability in large-scale wireless networks.  ...  networks to calibrate their inputs for a better performance.  ... 
arXiv:2108.01529v1 fatcat:5hio52yujnf33hjdvu6ca2buxa

Trainable Calibration Measures For Neural Networks From Kernel Mean Embeddings

Aviral Kumar, Sunita Sarawagi, Ujjwal Jain
2018 International Conference on Machine Learning  
Modern neural networks have recently been found to be poorly calibrated, primarily in the direction of over-confidence.  ...  Methods like entropy penalty and temperature smoothing improve calibration by clamping confidence, but in doing so compromise the many legitimately confident predictions.  ...  Acknowledgements We thank all anonymous reviewers for their comments and for pointing to the work on scoring rules in statistics.  ... 
dblp:conf/icml/KumarSJ18 fatcat:up2th6ojonempfue5rgexhugze

Application of Deep Learning Technique to an Analysis of Hard Scattering Processes at Colliders [article]

Lev Dudko, Petr Volkov, Georgii Vorotnikov, Andrei Zaborenko
2021 arXiv   pre-print
In this paper we will cover several methods of improving the performance of a deep neural network in a classification task in an instance of top quark analysis.  ...  Deep neural networks have rightfully won the place of one of the most accurate analysis tools in high energy physics.  ...  Conclusion We have demonstrated several approaches to improve the accuracy of classification based on a model of a Deep Neural Network in High Energy Physics.  ... 
arXiv:2109.08520v1 fatcat:khtsqtbbfrdxtn6t73g7sfingi

Towards Photorealistic Reconstruction of Highly Multiplexed Lensless Images

Salman Siddique Khan, Adarsh V R, Vivek Boominathan, Jasper Tan, Ashok Veeraraghavan, Kaushik Mitra
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
used for distributed monitoring.  ...  Our approach overcomes these drawbacks using a fully trainable non-iterative deep learning based model.  ...  One way to improve the reconstruction performance would be to exploit the natural image statistics within the data using data-driven techniques like convolutional neural networks [23] .  ... 
doi:10.1109/iccv.2019.00795 dblp:conf/iccv/KhanRBTVM19 fatcat:bxpcyjbpvbfplixzjgh354cisy

A compact butterfly-style silicon photonic-electronic neural chip for hardware-efficient deep learning [article]

Chenghao Feng, Jiaqi Gu, Hanqing Zhu, Zhoufeng Ying, Zheng Zhao, David Z. Pan, Ray T. Chen
2022 arXiv   pre-print
The optical neural network (ONN) is a promising hardware platform for next-generation neurocomputing due to its high parallelism, low latency, and low energy consumption.  ...  Here, we move beyond classical GEMM-based ONNs and propose an optical subspace neural network (OSNN) architecture, which trades the universality of weight representation for lower optical component usage  ...  We refer to such NN architectures as subspace neural networks.  ... 
arXiv:2111.06705v2 fatcat:6l3xfbxbtzbntedj7fdifhcuyu

Convolutional Neural Networks for Pose Recognition in Binary Omni-directional Images [chapter]

S. V. Georgakopoulos, K. Kottari, K. Delibasis, V. P. Plagianakos, I. Maglogiannis
2016 IFIP Advances in Information and Communication Technology  
Very recently, methods that use deep neural networks in order to tackle the problem of human pose estimation have started to appear in the literature.  ...  CNNs are trainable multistage architectures that belong to the first approach of classification methods [15] .  ... 
doi:10.1007/978-3-319-44944-9_10 fatcat:bwilgihpkjgx5dtvrj6gq7mwkm

Neural Clamping: Joint Input Perturbation and Temperature Scaling for Neural Network Calibration [article]

Yung-Chen Tang, Pin-Yu Chen, Tsung-Yi Ho
2022 arXiv   pre-print
calibration methods.  ...  Neural network calibration is an essential task in deep learning to ensure consistency between the confidence of model prediction and the true correctness likelihood.  ...  However, we aim to provide a new perspective and show that joint input-output model calibration can further improve neural network calibration.  ... 
arXiv:2209.11604v1 fatcat:67d6ltqirjdetjdzet2uue5yl4

Universal uncertainty estimation for nuclear detector signals with neural networks and ensemble learning [article]

Pengcheng Ai, Zhi Deng, Yi Wang, Chendi Shen
2022 arXiv   pre-print
Furthermore, ensemble learning is utilized to estimate the uncertainty originated from trainable parameters of the network and improve the robustness of the whole model.  ...  In this paper, we propose using multi-layer convolutional neural networks for empirical uncertainty estimation and feature extraction of nuclear pulse signals.  ...  Normalized calibration plots for comparison of traditional methods and neural networks in the experiment. The numbers in the brackets are A-UCE scores. Figure 10 .Figure 11 . 1011 Figure 10.  ... 
arXiv:2110.04975v3 fatcat:djvnpmf4rjg4bopbeejphzpjbm

Fully Dense Neural Network for the Automatic Modulation Recognition [article]

Miao Du, Qin Yu, Shaomin Fei, Chen Wang, Xiaofeng Gong, Ruisen Luo
2019 arXiv   pre-print
proposes a new network structure called Fully Dense Neural Network (FDNN).  ...  Nowadays, we mainly use various convolution neural network (CNN) structures to extract features from radio data or spectrogram in AMR.  ...  Convolution neural networks are a commonly used neural network for extracting features.  ... 
arXiv:1912.03449v1 fatcat:oiu4bebpyfhzle7e2ayagiercu

Soft Calibration Objectives for Neural Networks [article]

Archit Karandikar, Nicholas Cain, Dustin Tran, Balaji Lakshminarayanan, Jonathon Shlens, Michael C. Mozer, Becca Roelofs
2021 arXiv   pre-print
However, deep neural networks are often under- or over-confident in their predictions.  ...  Consequently, methods have been developed to improve the calibration of their predictive uncertainty both during training and post-hoc.  ...  Acknowledgements The authors thank Brennan McConnell and Mohammad Khajah who conducted initial explorations of soft binning calibration loss.  ... 
arXiv:2108.00106v2 fatcat:oxgu3qitdzb55b7lehp5ces5ii

Learned reconstructions for practical mask-based lensless imaging

Kristina Monakhova, Joshua Yurtsever, Grace Kuo, Nick Antipa, Kyrollos Yanny, Laura Waller
2019 Optics Express  
In this work, we address these limitations using a bounded-compute, trainable neural network to reconstruct the image.  ...  Finally, we test our network on images taken in the wild with a prototype mask-based camera, demonstrating that our network generalizes to natural images.  ...  Conclusion We presented several unrolled, model-based neural networks for lensless imaging with a varying number of trainable parameters.  ... 
doi:10.1364/oe.27.028075 fatcat:cu7a2uvw7zhv3hgd6iq3lrqige

Automated evaluation of Tuberculosis using Deep Neural Networks

Truong-Minh Le, Bao-Thien Nguyen-Tat, Vuong M. Ngo
2022 EAI Endorsed Transactions on Industrial Networks and Intelligent Systems  
of the neural networks.  ...  In terms of Deep Neural Networks, we picked up VGG16 as the baseline network architecture, then use other ones which are state-of-the-art networks for comparison purposes.  ...  For completing this project, we have collected the standard datasets as well as applied the most advanced neural network models to improve the confidence of models.  ... 
doi:10.4108/eetinis.v8i30.478 fatcat:vly5wevdevenxbkbejrioxedkq
« Previous Showing results 1 — 15 out of 4,221 results