1,711 Hits in 5.7 sec

Using Task Descriptions in Lifelong Machine Learning for Improved Performance and Zero-Shot Transfer [article]

David Isele, Mohammad Rostami, Eric Eaton
2017 arXiv   pre-print
Given only the descriptor for a new task, the lifelong learner is also able to accurately predict a model for the new task through zero-shot learning using the coupled dictionary, eliminating the need  ...  To reduce this burden, we develop a lifelong learning method based on coupled dictionary learning that utilizes high-level task descriptions to model the inter-task relationships.  ...  learning and zero-shot learning.  ... 
arXiv:1710.03850v1 fatcat:y42nqcdh4zah5csz5fmp2ylkpe

Using Task Descriptions in Lifelong Machine Learning for Improved Performance and Zero-Shot Transfer

Mohammad Rostami, David Isele, Eric Eaton
2020 The Journal of Artificial Intelligence Research  
Given only the descriptor for a new task, the lifelong learner is also able to accurately predict a model for the new task through zero-shot learning using the coupled dictionary, eliminating the need  ...  To reduce this burden, we develop a lifelong learning method based on coupled dictionary learning that utilizes high-level task descriptions to model inter-task relationships.  ...  Acknowledgments This research was supported by ONR grant #N00014-11-1-0139, AFRL grant #FA8750-14-1-0069, AFRL grant #FA8750-16-1-0109, and the DARPA Lifelong Learning Machines program under grant #FA8750  ... 
doi:10.1613/jair.1.11304 fatcat:gjbed6fp5jgaxpusimdxutanwi

Continuous Coordination As a Realistic Scenario for Lifelong Learning [article]

Hadi Nekoei, Akilesh Badrinaaraayanan, Aaron Courville, Sarath Chandar
2021 arXiv   pre-print
In this work, we introduce a multi-agent lifelong learning testbed that supports both zero-shot and few-shot settings.  ...  Current deep reinforcement learning (RL) algorithms are still highly task-specific and lack the ability to generalize to new environments.  ...  We can observe from Figure 3 and Table 1 that online EWC with Adam has the best average score in both the zero-shot and few-shot setting among the LLL algorithms, while the forgetting is least for  ... 
arXiv:2103.03216v2 fatcat:czld45ilrjfaxldxzmzcndfy5e

An Introduction to Advanced Machine Learning : Meta Learning Algorithms, Applications and Promises [article]

Farid Ghareh Mohammadi, M. Hadi Amini, Hamid R. Arabnia
2019 arXiv   pre-print
Machine learning employs a strict model or embedded engine to train and predict which still fails to learn unseen classes and sufficiently use online data.  ...  These algorithms, however, are not tailored for solving emerging learning problems. One of the important issues caused by online data is lack of sufficient samples per class.  ...  Information system Figure7 provides a general view of few shot learning, one-shot learning and 2-shot learning and generalized k-shot learning.  ... 
arXiv:1908.09788v1 fatcat:qujten7zzzbd7laazhymnfw2yi

Cross-lingual Lifelong Learning [article]

Meryem M'hamdi, Xiang Ren, Jonathan May
2022 arXiv   pre-print
In this paper, we present the Cross-lingual Lifelong Learning (CLL) challenge, where a model is continually fine-tuned to adapt to emerging data from different languages.  ...  To surmount such challenges, we benchmark a representative set of cross-lingual continual learning algorithms and analyze their knowledge preservation, accumulation, and generalization capabilities compared  ...  To analyze the zero-shot generalization to unseen languages, we analyze the performance of each model across different hops.  ... 
arXiv:2205.11152v1 fatcat:r7f6zlikdjhalfoa7vwkb7crlq

PRIMAL2: Pathfinding via Reinforcement and Imitation Multi-Agent Learning – Lifelong [article]

Mehul Damani and Zhiyao Luo and Emerson Wenzel and Guillaume Sartoretti
2021 arXiv   pre-print
agents learn fully decentralized policies to reactively plan paths online in a partially observable world.  ...  In particular, this work addresses lifelong MAPF (LMAPF) - an online variant of the problem where agents are immediately assigned a new goal upon reaching their current one - in dense and highly structured  ...  We consider two variants of MAPF: one-shot MAPF and lifelong MAPF (LMAPF).  ... 
arXiv:2010.08184v2 fatcat:nmmg3mz74veurjxmflzp546aum

Learn Continually, Generalize Rapidly: Lifelong Knowledge Accumulation for Few-shot Learning [article]

Xisen Jin, Bill Yuchen Lin, Mohammad Rostami, Xiang Ren
2021 arXiv   pre-print
Existing models that pursue rapid generalization to new tasks (e.g., few-shot learning methods), however, are mostly trained in a single shot on fixed datasets, unable to dynamically expand their knowledge  ...  We present a new learning setup, Continual Learning of Few-Shot Learners (CLIF), to address the challenges of both learning settings in a unified setup.  ...  Episodic memory in lifelong language learning.  ... 
arXiv:2104.08808v3 fatcat:q4gs4qw6uzgbda2hkv7oju53di

Recent Advances in Zero-shot Recognition [article]

Yanwei Fu, Tao Xiang, Yu-Gang Jiang, Xiangyang Xue, Leonid Sigal, and Shaogang Gong
2017 arXiv   pre-print
One approach to scaling up the recognition is to develop models capable of recognizing unseen categories without any training instances, or zero-shot recognition/ learning.  ...  or when zero-shot recognition is implemented in a real-world setting.  ...  Yanwei Fu is supported by The Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning.  ... 
arXiv:1710.04837v1 fatcat:u3mp6dgj2rgqrarjm4dcywegmy

Lifelong Learning in Sensor-Based Human Activity Recognition

Juan Ye
2019 2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)  
Driven by these challenges, in this article we argue the need to move beyond learning to lifelong machine learning -with the ability to incrementally and continuously adapt to changes in the environment  ...  We introduce a conceptual framework for lifelong machine learning to structure various relevant proposals in the area, and identify some key research challenges that remain.  ...  Zero-Shot Learning. Learning a model for new activities often suffers from the scarcity of training data.  ... 
doi:10.1109/percomw.2019.8730783 dblp:conf/percom/Ye19 fatcat:6hytlwr54fhddazculsyzs4yni

Lifelong Machine Learning Architecture for Classification

Xianbin Hong, Sheng-Uei Guan, Ka Lok Man, Prudence W. H. Wong
2020 Symmetry  
Such a learning architecture also is commonly called as lifelong machine learning (LML).  ...  In order to demonstrate the application of lifelong machine learning, this paper proposed a novel and symmetric lifelong learning approach for sentiment classification as an example to show how it adapts  ...  This also is the foundation of the one shot learning or even zero shot learning.  ... 
doi:10.3390/sym12050852 fatcat:4oranwahyndwhfexvdt2w7d2eu

Exploring the Challenges towards Lifelong Fact Learning [article]

Mohamed Elhoseiny, Francesca Babiloni, Rahaf Aljundi, Marcus Rohrbach, Manohar Paluri, Tinne Tuytelaars
2018 arXiv   pre-print
So far life-long learning (LLL) has been studied in relatively small-scale and relatively artificial setups. Here, we introduce a new large-scale alternative.  ...  We named that evaluation Generalized lifelong learning (G-LLL), in line with the idea of Generalized zero-shot learning proposed in [3] . We detail the evaluation metric in Sec. 5.1.  ...  Fig. 1 . 1 Lifelong Fact Learning Fig. 2.  ... 
arXiv:1812.10524v1 fatcat:eorgzymzsjeabh7fg7c53vuhra

Controlled Forgetting: Targeted Stimulation and Dopaminergic Plasticity Modulation for Unsupervised Lifelong Learning in Spiking Neural Networks [article]

Jason M. Allred, Kaushik Roy
2019 arXiv   pre-print
For a deployed system that must learn online from an uncontrolled and unknown environment, the ordering of input samples often fails to meet this criterion, making lifelong learning a difficult challenge  ...  This targeting controls the forgetting process in a way that reduces the degradation of accuracy for older tasks while learning new tasks.  ...  The online CFN shows comparable accuracy while successfully achieving lifelong learning with the tasks completely disjoint and sequential.  ... 
arXiv:1902.03187v2 fatcat:x73cjkltz5hwlkzhwmmhw2ketq

Recent Advances in Transfer Learning for Cross-Dataset Visual Recognition: A Problem-Oriented Perspective [article]

Jing Zhang and Wanqing Li and Philip Ogunbona and Dong Xu
2019 arXiv   pre-print
The comprehensive problem-oriented review of the advances in transfer learning with respect to the problem has not only revealed the challenges in transfer learning for visual recognition, but also the  ...  This paper takes a problem-oriented perspective and presents a comprehensive review of transfer learning methods, both shallow and deep, for cross-dataset visual recognition.  ...  sequential/online transfer learning, and closely related to lifelong learning [141, 192, 218] .  ... 
arXiv:1705.04396v3 fatcat:iknfmppi5zca7ljovdlwvdwluu

Towards a theory of out-of-distribution learning [article]

Ali Geisa, Ronak Mehta, Hayden S. Helm, Jayanta Dey, Eric Eaton, Jeffery Dick, Carey E. Priebe, Joshua T. Vogelstein
2022 arXiv   pre-print
We then define and prove the relationship between generalized notions of learnability, and show how this framework is sufficiently general to characterize transfer, multitask, meta, continual, and lifelong  ...  What is learning?  ...  Third, we do not require large sample sizes, rather, our definition includes zero-shot and few-shot learning.  ... 
arXiv:2109.14501v4 fatcat:nd5ffokjybhipcyygbtakdw4ke

Continual Learning of Recurrent Neural Networks by Locally Aligning Distributed Representations [article]

Alexander Ororbia, Ankur Mali, C. Lee Giles, Daniel Kifer
2019 arXiv   pre-print
We present results that show the P-TNCN's ability to conduct zero-shot adaptation and online continual sequence modeling.  ...  Significantly, the hidden unit correction phase of P-TNCN allows it to adapt to new datasets even if its synaptic weights are held fixed (zero-shot adaptation) and facilitates retention of prior generative  ...  APPENDIX In this appendix, we present the qualitative exploration of the P-TNCN's ability to process out-of-domain inputs as well as an extra experiment in zero-shot adaptation.  ... 
arXiv:1810.07411v4 fatcat:feda6a4tkjeznieq674vbic27m
« Previous Showing results 1 — 15 out of 1,711 results