A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Filters
Novel Class Discovery without Forgetting
[article]
2022
arXiv
pre-print
Preserved Fulltext
Inspired by this, we identify and formulate a new, pragmatic problem setting of NCDwF: Novel Class Discovery without Forgetting, which tasks a machine learning model to incrementally discover novel categories ...
enhances unsupervised discovery of novel classes, and 3) a simple Known Class Identifier which aids generalized inference when the testing data contains instances form both seen and unseen categories. ...
Novel Class Discovery without Forgetting We formally define Novel Class Discovery without Forgetting in Sec. ...
arXiv:2207.10659v1
fatcat:ri5rafx37feqzlzc74jwj5tp5q
Proxy Anchor-based Unsupervised Learning for Continuous Generalized Category Discovery
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2307.10943v1 fatcat:4uzyal7h5jhl7p7snudtu3napu
sets without prior knowledge. ...
Existing methods for novel category discovery are limited by their reliance on labeled datasets and prior knowledge about the number of novel categories and the proportion of novel samples in the batch ...
As shown in Table 5 , the method without the exemplar recorded the highest novel discovery performances but also showed the highest forgetting. ...
arXiv:2307.10943v2
fatcat:vssx6ynrpbbjrb5vzdxzivn2cy
Large-scale Pre-trained Models are Surprisingly Strong in Incremental Novel Class Discovery
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Other Versions
for a limited number of incremental steps (e.g., class-iNCD). ...
Discovering novel concepts from unlabelled data and in a continuous manner is an important desideratum of lifelong learners. ...
Pre.) on the novel class discovery performance. ...
arXiv:2303.15975v2
fatcat:ozinnfgaerbgrpl2v5xqndokxi
Grow and Merge: A Unified Framework for Continuous Categories Discovery
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
A common challenge faced by novel category discovery is that different sets of features are needed for classification and category discovery: class discriminative features are preferred for classification ...
Although a number of studies are devoted to novel category discovery, most of them assume a static setting where both labeled and unlabeled data are given at once for finding new categories. ...
The maximum forgetting M f and final discovery M d are reported in Table 2 . ...
arXiv:2210.04174v1
fatcat:ri634va4mfdmhdgqib7r7ajaue
Class-incremental Novel Class Discovery
2023
Zenodo
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
We study the new task of class-incremental Novel Class Discovery (class-iNCD), which refers to the problem of discovering novel categories in an unlabelled data set by leveraging a pre-trained model that ...
Inspired by rehearsal-based incremental learning methods, in this paper we propose a novel approach for class-iNCD which prevents forgetting of past information about the base classes by jointly exploiting ...
Contrarily in class-iNCD, as the novel classes need to be learned without explicit supervision, it makes the optimization of NCD part interfere with that of forgetting. ...
doi:10.5281/zenodo.7566120
fatcat:w77svmshpjfppm7ywx3uglrqbq
Class-incremental Novel Class Discovery
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
We study the new task of class-incremental Novel Class Discovery (class-iNCD), which refers to the problem of discovering novel categories in an unlabelled data set by leveraging a pre-trained model that ...
Inspired by rehearsal-based incremental learning methods, in this paper we propose a novel approach for class-iNCD which prevents forgetting of past information about the base classes by jointly exploiting ...
A1 Dataset statistics for class-incremental novel class discovery. ...
arXiv:2207.08605v1
fatcat:yyyop5xa2bgcjg5gyojda5nb2q
Page 50 of Belgravia Vol. 16, Issue
[page]
1871
Belgravia
Preserved Fulltext
The Internet Archive has digitized a microfilm copy of this work. It may be possible to borrow a copy for reading.
The problem of suffering and sorrow is that which lies at the root of all novels: it forms the interest and pathos of every life; it is the fons et origo lacrimarum ; without it there would be no romance ...
, for there would be no doubt; and where the milk-and-water novelist takes infinit- esimal pinches, creating imaginary sorrows out of fictitious sins—a heroine laughs in church, or forgets to tell her ...
Towards Open-Set Object Detection and Discovery
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
With this method, a detector is able to detect objects belonging to known classes and define novel categories for objects of unknown classes with minimal supervision. ...
classes. ...
The model uses example replay to make the open-set detector learn new classes incrementally without forgetting the previous ones. ...
arXiv:2204.05604v1
fatcat:adntvubdmjg3nfwirbvzy6idiu
Exemplar-Based Contrastive Self-Supervised Learning with Few-Shot Class Incremental Learning
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
A major benefit of the extensions is that exemplar-based CSSL, with supervised finetuning, supports few-shot class incremental learning (CIL). ...
(For unlabeled samples, their class exemplars are nearest-neighbors, with or without a class label.)
3. 1 NCL 1 NCL (Neighborhood Contrastive Learning) [4] addresses Novel Class Discovery (NCD), the ...
However, it is desirable to leverage the labeled data (collected from known classes) to explore the unlabeled data for discovery of novel classes. ...
arXiv:2202.02601v1
fatcat:bwhivxbocraonlmxh25xp2uuea
Towards Label-Efficient Incremental Learning: A Survey
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Other Versions
To that end, researchers study incremental learning, where a learner is required to adapt to an incoming stream of data with a varying distribution while preventing forgetting of past knowledge. ...
Finally, we identify novel directions that can further enhance label-efficiency and improve incremental learning scalability. Project website: https://github.com/kilickaya/label-efficient-il. ...
The emphasis of CEC is more on the novel class learning rather than maintaining base class performance, which eventually exacerbates forgetfulness.
Clustering-Based Methods IDL-VQ. ...
arXiv:2302.00353v3
fatcat:vj6h4utk5fdzlpas7balsmrpaq
Incremental Generalized Category Discovery
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2304.14310v1 fatcat:hh2xnzpkmrclhnsr5nppfxfzfeOther Versions
We explore the problem of Incremental Generalized Category Discovery (IGCD). ...
We present a new method for IGCD which combines non-parametric categorization with efficient image sampling to mitigate catastrophic forgetting. ...
This obtains a good balance between forgetting and discovery. ...
arXiv:2304.14310v3
fatcat:qtydiepgwvbb5ic7ho77iknaiu
Non-Exemplar Online Class-incremental Continual Learning via Dual-prototype Self-augment and Refinement
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2303.10891v1 fatcat:fgzzbadryfeqnnkpgijtdzttfuOther Versions
The challenges of this task are mainly two-fold: (1) Both base and novel classes suffer from severe catastrophic forgetting as no previous samples are available for replay. (2) As the online data can only ...
This paper investigates a new, practical, but challenging problem named Non-exemplar Online Class-incremental continual Learning (NO-CL), which aims to preserve the discernibility of base classes without ...
Novel Class Discovery (C-NCD) [21, 49] , where base classes are well trained, and the knowledge are retained, but novel classes need to be explored. ...
arXiv:2303.10891v3
fatcat:34yxpzqcpfairfer74zo3mfr3m
Automatically Discovering and Learning New Visual Categories with Ranking Statistics
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
We tackle the problem of discovering novel classes in an image collection given labelled examples of other classes. ...
We evaluate our approach on standard classification benchmarks and outperform current methods for novel category discovery by a significant margin. ...
Methods for novel category discovery such as Hsu et al., 2019; focus on obtaining the highest clustering accuracy for the new unlabelled classes, but may forget the existing labelled classes in the process ...
arXiv:2002.05714v1
fatcat:iwcciudgyrgulkdn6nyh7zugfa
Class-relation Knowledge Distillation for Novel Class Discovery
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2307.09158v1 fatcat:hicbv4elrzcfjavm3yncfumftuOther Versions
We tackle the problem of novel class discovery, which aims to learn novel classes without supervision based on labeled data from known classes. ...
Empirically, we find that such class relation becomes less informative during typical discovery training. ...
Discussion with NCDwF [18] In NCDwF, they focus on novel class discovery without forgetting, where known class data is not available in the discovery stage. ...
arXiv:2307.09158v3
fatcat:nlxlval6cfcurpzqx7noiimcg4
Stream-51: Streaming Classification and Novelty Detection from Videos
2020
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Once trained and deployed in a real-time environment, these models struggle to identify novel inputs not initially represented in the training distribution. ...
Further, they cannot be easily updated on new information or they will catastrophically forget previously learned knowledge. ...
In general, the models that were fine-tuned without a buffer performed poorly since they did not have any mechanisms to mitigate forgetting. ...
doi:10.1109/cvprw50498.2020.00122
dblp:conf/cvpr/RoadyHVK20
fatcat:a5i4ko6ojvgahdaiwxs2sgjvai
« Previous
Showing results 1 — 15 out of 37,757 results