Unsupervised Outlier Detection using Memory and Contrastive Learning

Dou Quan; Jocelyn Chanussot; Licheng Jiao; Ning Huyan; Xiangrong Zhang; Xuefeng Liang

RECORD

Unsupervised Outlier Detection using Memory and Contrastive Learning

Ning Huyan, Dou Quan, Xiangrong Zhang, Xuefeng Liang, Jocelyn Chanussot, Licheng Jiao

Computer Vision and Pattern Recognition · Submitted 27 Jul 2021

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim Extensive experiments on four benchmark datasets show that our proposed MCOD achieves a considerable performance and outperforms nine state-of-the-art methods.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2107.12642 (version 1)

Pith IDGHDHT2WH, Pith's permanent record of this paper and its review

CitationDou Quan, Jocelyn Chanussot, Licheng Jiao, Ning Huyan, Xiangrong Zhang, Xuefeng Liang. "Unsupervised Outlier Detection using Memory and Contrastive Learning." arXiv:2107.12642 (2021). https://pith.science/paper/GHDHT2WH

Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 6,734 input tokens.

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.09 · human participants 0.03 · controlled comparison 0.47 · randomized assignment 0.01 · states quantitative result 0.39 · compares with prior results 0.97 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

Outlier detection is one of the most important processes taken to create good, reliable data in machine learning. The most methods of outlier detection leverage an auxiliary reconstruction task by assuming that outliers are more difficult to be recovered than normal samples (inliers). However, it is not always true, especially for auto-encoder (AE) based models. They may recover certain outliers even outliers are not in the training data, because they do not constrain the feature learning. Instead, we think outlier detection can be done in the feature space by measuring the feature distance between outliers and inliers. We then propose a framework, MCOD, using a memory module and a contrastive learning module. The memory module constrains the consistency of features, which represent the normal data. The contrastive learning module learns more discriminating features, which boosts the distinction between outliers and inliers. Extensive experiments on four benchmark datasets show that our proposed MCOD achieves a considerable performance and outperforms nine state-of-the-art methods.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	Outlier detection is one of the most important processes taken to create good, reliable data in machine learning.	background 1.00	0.04	0.02	asserted 1.00
s1	The most methods of outlier detection leverage an auxiliary reconstruction task by assuming that outliers are more difficult to be recovered than normal samples (inliers).	background 1.00	0.04	0.03	asserted 0.98
s2	However, it is not always true, especially for auto-encoder (AE) based models.	gap 0.66	0.09	0.03	asserted 0.86
s3	They may recover certain outliers even outliers are not in the training data, because they do not constrain the feature learning.	background 0.60	0.10	0.04	hedged possibility 0.99
s4	Instead, we think outlier detection can be done in the feature space by measuring the feature distance between outliers and inliers.	method 0.45	0.21	0.19	asserted 0.41
s5	We then propose a framework, MCOD, using a memory module and a contrastive learning module.	contribution 0.80	0.90	0.11	asserted 1.00
s6	The memory module constrains the consistency of features, which represent the normal data.	method 0.97	0.65	0.06	asserted 1.00
s7	The contrastive learning module learns more discriminating features, which boosts the distinction between outliers and inliers.	method 0.97	0.74	0.10	asserted 1.00
s8	Extensive experiments on four benchmark datasets show that our proposed MCOD achieves a considerable performance and outperforms nine state-of-the-art methods.	result 1.00	0.96	0.80	asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.95.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

Outlier detection is one of the most important processes taken to create good, reliable data in machine learning. The most methods of outlier detection leverage an auxiliary reconstruction task by assuming that outliers are more difficult to be recovered than normal samples (inliers). However, it is not always true, especially for auto-encoder (AE) based models. They may recover certain outliers even outliers are not in the training data, because they do not constrain the feature learning. Instead, we think outlier detection can be done in the feature space by measuring the feature distance between outliers and inliers. We then propose a framework, MCOD, using a memory module and a contrastive learning module. The memory module constrains the consistency of features, which represent the normal data. The contrastive learning module learns more discriminating features, which boosts the distinction between outliers and inliers. Extensive experiments on four benchmark datasets show that our proposed MCOD achieves a considerable performance and outperforms nine state-of-the-art methods.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

4 arXiv papers cite this in 6 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 2 sentences without a confident label

Cites it as support 2 papers

In addition, once there exist abnormal instances in the training data, the performance of these OCC-based detection methods could be severely degraded [ref] .

Detecting Multivariate Time Series Anomalies with Zero Known Label · Time Series Anomaly Detection · 0.97
As shown in Fig. [ref], this degrades the performance of those OCC-based methods [ref] .

Label-Free Multivariate Time Series Anomaly Detection · Introduction · 0.98

What it cites · 6 sentences without a confident label

Uses the method or tool of 4 papers

To optimize the embedding representations from contrastive learning, we adopt the InfoNCE loss [ref]

Representation Learning with Contrastive Predictive Coding · Contrastive learning for discriminating feature · 1.00
To optimize the embedding representations from contrastive learning, we adopt the InfoNCE loss [ref]

Improved Baselines with Momentum Contrastive Learning · Contrastive learning for discriminating feature · 1.00
Moreover, we add a regularization to prevent the model falling into a local minimum where the majority gathers to the minority clusters [ref] .

Deep Robust Clustering by Contrastive Learning · Memory module for consistent feature · 0.86

Uses data or values from 1 paper

The experiments are conducted on four benchmarks: MNIST [ref] , Fashion-MNIST [ref] , CIFAR10 [ref] , and CIFAR100 [ref] .

Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms · Experiment · 0.97

Compares with 2 papers

To demonstrate the superiority of our MCOD, we compare the MCOD with the baselines OC-SVM [ref] and nine state-of-the-arts including CAE [ref] , DRAE [ref] , DSEBM [ref] , RDAE [ref] , MOGAL [ref] , RSRAE [ref] , RSRAE+ [ref] SSD-IF [ref] and E3Out [ref] .

Deep Structured Energy Based Models for Anomaly Detection · Experiment · 1.00
To demonstrate the superiority of our MCOD, we compare the MCOD with the baselines OC-SVM [ref] and nine state-of-the-arts including CAE [ref] , DRAE [ref] , DSEBM [ref] , RDAE [ref] , MOGAL [ref] , RSRAE [ref] , RSRAE+ [ref] SSD-IF [ref] and E3Out [ref] .

Robust Subspace Recovery Layer for Unsupervised Anomaly Detection · Experiment · 1.00

Cites as background 2 papers

The reconstruction-based OD methods often employ auto-encoders to detect outliers by measuring the reconstruction errors [ref] .

Robust Subspace Recovery Layer for Unsupervised Anomaly Detection · Related work · 0.95
The mainstream approaches of outlier detection (OD) are unsupervised frameworks, which assume that outliers are minority and significantly deviated from inliers [ref] .

Deep Learning for Anomaly Detection: A Survey · Related work · 0.91

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.