Filters








15,751 Hits in 5.7 sec

Agnostic learning with unknown utilities [article]

Kush Bhatia, Peter L. Bartlett, Anca D. Dragan, Jacob Steinhardt
2021 arXiv   pre-print
We formally study this as agnostic learning with unknown utilities: given a dataset S = {x_1, ..., x_n} where each data point x_i ∼𝒟, the objective of the learner is to output a function f in some class  ...  This risk measures the performance of the output predictor f with respect to the best predictor in the class ℱ on the unknown underlying utility u^*.  ...  We propose a novel framework, which we call agnostic learning with unknown utilities, for studying decision problems wherein the learner is evaluated with respect to an unknown utility function.  ... 
arXiv:2104.08482v1 fatcat:hxzvbvvjrbdufmlofu6cjdpkku

Agnostic Learning with Unknown Utilities

Kush Bhatia, Peter L. Bartlett, Anca D. Dragan, Jacob Steinhardt, James R. Lee
2021
We formally study this as agnostic learning with unknown utilities: given a dataset S = {x_1, ..., x_n} where each data point x_i ∼ 𝒟_x from some unknown distribution 𝒟_x, the objective of the learner  ...  This risk measures the performance of the output predictor f with respect to the best predictor in the class ℱ on the unknown underlying utility u^*:𝒳×𝒴↦ [0,1].  ...  mispredicting it I T C S 2 0 2 1 55:4 Agnostic Learning with Unknown Utilities as a red signal.  ... 
doi:10.4230/lipics.itcs.2021.55 fatcat:zfy365esqrcodhd4axeiyusj2y

Class-agnostic Object Detection with Multi-modal Transformer [article]

Muhammad Maaz, Hanoona Rasheed, Salman Khan, Fahad Shahbaz Khan, Rao Muhammad Anwer, Ming-Hsuan Yang
2022 arXiv   pre-print
Towards this goal, numerous learning-free and learning-based approaches have been developed to score objectness. However, they generally do not scale well across new domains and novel objects.  ...  For the first time in literature, we demonstrate that Multi-modal Vision Transformers (MViT) trained with aligned image-text pairs can effectively bridge this gap.  ...  It utilizes example-replay strategy [66] for alleviating forgetting, when progressively learning the unknown categories once their labels become available.  ... 
arXiv:2111.11430v6 fatcat:q6xf7mdrmzcaji7qytav7xkpta

Fairness-aware Agnostic Federated Learning [article]

Wei Du, Depeng Xu, Xintao Wu, Hanghang Tong
2020 arXiv   pre-print
In this paper, we develop a fairness-aware agnostic federated learning framework (AgnosticFair) to deal with the challenge of unknown testing distribution.  ...  Federated learning is an emerging framework that builds centralized machine learning models with training data distributed across multiple devices.  ...  The authors in [20] propose agnostic federated learning to deal with the unknown testing data distribution.  ... 
arXiv:2010.05057v1 fatcat:cr4zyopz6vbhflatdlt55opzdq

Class-agnostic Object Detection [article]

Ayush Jaiswal, Yue Wu, Pradeep Natarajan, Premkumar Natarajan
2020 arXiv   pre-print
However, due to the difficulty and cost associated with creating and annotating detection datasets, trained models detect a limited number of object types with unknown objects treated as background content  ...  Experimental results show that adversarial learning improves class-agnostic detection efficacy.  ...  Besides the problem formulation, we propose training and evaluation protocols for class-agnostic detection with generalization and downstream utility as primary goals.  ... 
arXiv:2011.14204v1 fatcat:gqlsiucnhbcklpmgyynwidszoa

Exploring Category-Agnostic Clusters for Open-Set Domain Adaptation [article]

Yingwei Pan and Ting Yao and Yehao Li and Chong-Wah Ngo and Tao Mei
2020 arXiv   pre-print
., unknown class). The extension of domain adaptation from closed-set to such open-set situation is not trivial since the target samples in unknown class are not expected to align with the source.  ...  Specifically, we present Self-Ensembling with Category-agnostic Clusters (SE-CC) -- a novel architecture that steers domain adaptation with the additional guidance of category-agnostic clusters that are  ...  Later on, [29] utilizes adversarial training to learn feature representations that could separate the target samples of unknown class from the known target samples.  ... 
arXiv:2006.06567v1 fatcat:plisr3t645exjk4m7dioq7awjm

Exploring Category-Agnostic Clusters for Open-Set Domain Adaptation

Yingwei Pan, Ting Yao, Yehao Li, Chong-Wah Ngo, Tao Mei
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Specifically, we present Self-Ensembling with Category-agnostic Clusters (SE-CC) -a novel architecture that steers domain adaptation with the additional guidance of category-agnostic clusters that are  ...  ., unknown class). The extension of domain adaptation from closedset to such open-set situation is not trivial since the target samples in unknown class are not expected to align with the source.  ...  Later on, [29] utilizes adversarial training to learn feature representations that could separate the target samples of unknown class from the known target samples.  ... 
doi:10.1109/cvpr42600.2020.01388 dblp:conf/cvpr/PanYLNM20 fatcat:azseeqfzanbj3oqlm2qwyqihri

TAPE: Task-Agnostic Prior Embedding for Image Restoration [article]

Lin Liu, Lingxi Xie, Xiaopeng Zhang, Shanxin Yuan, Xiangyu Chen, Wengang Zhou, Houqiang Li, Qi Tian
2022 arXiv   pre-print
Learning a generalized prior for natural image restoration is an important yet challenging task.  ...  Recently, deep neural networks have been used to learn various image priors but do not guarantee to generalize.  ...  Learning task-agnostic priors and pixel-wise contrastive loss on pre-training stage can help the performance of fine-tuning on the unknown tasks.  ... 
arXiv:2203.06074v2 fatcat:dblwhkbyr5fifccjpavxjgs2zm

Open-Source Tools and Containers for the Production of Large-Scale S/TEM Datasets

Alexander M Rakowski, Joydeep Munshi, Benjamin Savitzky, Shreyas Cholia, Matthew L Henderson, Maria KY Chan, Colin Ophus
2021 Microscopy and Microanalysis  
We show that the process is platform agnostic and scalable by producing examples on a single local machine and a high-performance computing (HPC) cluster with GPU acceleration.  ...  Such networks offer great promise and could ultimately, for example, perform strain and orientation mapping of materials with currently inaccessibly large fields of view, classification of unknown polycrystalline  ...  Such networks offer great promise and could ultimately, for example, perform strain and orientation mapping of materials with currently inaccessibly large fields of view, classification of unknown polycrystalline  ... 
doi:10.1017/s1431927621000829 fatcat:e7cqc7ddpfbsldb3wlaqe7pkl4

Monocular Instance Motion Segmentation for Autonomous Driving: KITTI InstanceMotSeg Dataset and Multi-task Baseline [article]

Eslam Mohamed, Mahmoud Ewaisha, Mennatullah Siam, Hazem Rashed, Senthil Yogamani, Waleed Hamdy, Muhammad Helmi, Ahmad El-Sallab
2021 arXiv   pre-print
The model then learns separate prototype coefficients within the class agnostic and semantic heads providing two independent paths of object detection for redundant safety.  ...  To obtain real-time performance, we study different efficient encoders and obtain 39 fps on a Titan Xp GPU using MobileNetV2 with an improvement of 10% mAP relative to the baseline.  ...  In this case we call it class agnostic segmentation as it is solely dependant on motion cues regardless of semantics which would scale better to unknown objects.  ... 
arXiv:2008.07008v4 fatcat:a54do7k7rrdhdj75dm6wyom5qy

Stream-based Active Learning with Verification Latency in Non-stationary Environments [article]

Andrea Castellani, Sebastian Schmitt, Barbara Hammer
2022 arXiv   pre-print
Data stream classification is an important problem in the field of machine learning.  ...  We propose PRopagate (PR), a latency independent utility estimator which also predicts the requested, but not yet known, labels.  ...  Conclusion and future work In this work, we addressed the problem of Active Learning (AL) under finite, variable, and unknown verification latency.  ... 
arXiv:2204.06822v1 fatcat:uirw2p7e5zcxvm42vxpsn47b7i

The Challenges of Exploration for Offline Reinforcement Learning [article]

Nathan Lambert, Markus Wulfmeier, William Whitney, Arunkumar Byravan, Michael Bloesch, Vibhavari Dasagi, Tim Hertweck, Martin Riedmiller
2022 arXiv   pre-print
With Explore2Offline, we propose to evaluate the quality of collected data by transferring the collected data and inferring policies with reward relabelling and standard offline RL algorithms.  ...  Offline Reinforcement Learning (ORL) enablesus to separately study the two interlinked processes of reinforcement learning: collecting informative experience and inferring optimal behaviour.  ...  Conclusion We introduce Explore2Offline, a method for utilizing taskagnostic data for policy learning of unknown downstream tasks.  ... 
arXiv:2201.11861v2 fatcat:sajzyrnxuze6lo2lozj4szy4um

Large-Scale, Language-Agnostic Discourse Classification of Tweets During COVID-19

Oguzhan Gencoglu
2020 Machine Learning and Knowledge Extraction  
For this purpose, we propose language-agnostic tweet representations to perform large-scale Twitter discourse classification with machine learning.  ...  utilization of these representations.  ...  We utilize state-of-the-art language-agnostic tweet representations coupled with simple, lightweight classifiers to be able to capture COVID-19 related discourse during a span of 13 weeks.  ... 
doi:10.3390/make2040032 fatcat:z5yyzynqizefxhgov7dcktpyku

Blind and Channel-agnostic Equalization Using Adversarial Networks [article]

Vincent Lauinger, Manuel Hoffmann, Jonas Ney, Norbert Wehn, Laurent Schmalen
2022 arXiv   pre-print
The learning is only based on the statistics of the transmit signal, so it is blind regarding the actual transmit symbols and agnostic to the channel model.  ...  To tackle those challenges, we propose a novel adaptive equalization scheme, which exploits the prosperous advances in deep learning by training an equalizer with an adversarial network.  ...  Channel-agnostic algorithms, however, can also be used if the channel model is unknown or if it deviates significantly from the real channel.  ... 
arXiv:2209.07277v1 fatcat:nkku5rcjubdkdmsztgvthnc7ei

Task-agnostic Continual Hippocampus Segmentation for Smooth Population Shifts [article]

Camila Gonzalez, Amin Ranem, Ahmed Othman, Anirban Mukhopadhyay
2022 arXiv   pre-print
We explore how such methods perform in a task-agnostic setting that more closely resembles dynamic clinical environments with gradual population shifts.  ...  We propose ODEx, a holistic solution that combines out-of-distribution detection with continual learning techniques.  ...  In task-agnostic scenarios, task labels j are unknown and may not even be clearly defined.  ... 
arXiv:2208.03206v1 fatcat:dr4m5mrd6fazrfovufk6hqgz7m
« Previous Showing results 1 — 15 out of 15,751 results