Time series feature learning with labeled and unlabeled data

Wang, Haishuai; Zhang, Qin; Wu, Jia; Pan, Shirui; Chen, Yixin

doi:10.1016/j.patcog.2018.12.026

Cited by 108 publications

(60 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since the data generating processes are completely different 1 , the proposed method's performance can be judged without bias to similar data generating processes. Previously proposed methods were compared on the same in [17]. A summary of the datasets is given in Table 1.…”

Section: Methodsmentioning

confidence: 99%

“…The authors of [22] proposed a graph theoretic SSL algorithm that constructs graphs relating all samples based on different distance functions and consequently propagates labels. The current state-of-the-art method in the field [17] is based on shapelet learning [8] on both labeled and unlabeled time series data. On the other hand, we note recent works [4,5,7,15,23] which showed that strong supervision could be leveraged by describing a task that is inherent in the data itself (requires no manual annotation).…”

Section: Related Workmentioning

confidence: 99%

“…With motivations behind our method set from the literature review, we now draw the following insights: firstly there exist semi-supervised learning approaches similar to ours that learn from unlabeled data, most notably current state-of-the-art shapelet learning approach [17] if we consider shapelets to be similar to convolutional filters. However, with our work we exploit deep learning based methods which solve an auxiliary self-supervised task of forecasting which forces the network to learn filters to solve this particular complex task.…”

Section: Related Workmentioning

confidence: 99%

“…SSSL [17] is the current state-of-the-art method in the field. It uses shapelets to classify unlabeled samples thereby producing pseudo-labels.…”

Section: Dtw-d [2]mentioning

confidence: 99%

“…In reality, however, there is a high cost associated in acquiring such labeled data. As a result, there have been efforts to utilize semi-supervised learning algorithms catered especially for time series classification [2,9,12,17,21,22]. The idea behind semi-supervised learning is to exploit unlabeled data for training purpose in the presence of only few labeled instances.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Self-supervised Learning for Semi-supervised Time Series Classification

Jawed

Grabocka

Schmidt-Thieme

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Self-supervised learning is a promising new technique for learning representative features in the absence of manual annotations. It is particularly efficient in cases where labeling the training data is expensive and tedious, naturally linking it to the semi-supervised learning paradigm. In this work, we propose a new semi-supervised time series classification model that leverages features learned from the selfsupervised task on unlabeled data. The idea is to exploit the unlabeled training data with a forecasting task which provides a strong surrogate supervision signal for feature learning. We draw from established multitask learning approaches and model forecasting as an auxiliary task to be optimized jointly with the main task of classification. We evaluate our proposed method on benchmark time series classification datasets in semi-supervised setting and are able to show that it significantly outperforms the state-of-the-art baselines.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

“…SSSL [17] is the current state-of-the-art method in the field. It uses shapelets to classify unlabeled samples thereby producing pseudo-labels.…”

Section: Dtw-d [2]mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Self-supervised Learning for Semi-supervised Time Series Classification

Jawed

Grabocka

Schmidt-Thieme

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

SelfMatch: Robust semisupervised time‐series classification with self‐distillation

Xing

Xiao

Zhan

et al. 2022

Int J of Intelligent Sys

View full text Add to dashboard Cite

Over the years, a number of semisupervised deeplearning algorithms have been proposed for timeseries classification (TSC). In semisupervised deep learning, from the point of view of representation hierarchy, semantic information extracted from lower levels is the basis of that extracted from higher levels. The authors wonder if high-level semantic information extracted is also helpful for capturing low-level semantic information. This paper studies this problem and proposes a robust semisupervised model with self-distillation (SD) that simplifies existing semisupervised learning (SSL) techniques for TSC, called SelfMatch. SelfMatch hybridizes supervised learning, unsupervised learning, and SD.

show abstract

Port Performance Measures in Brazil: An Analysis in Port of Santos

Santos

Reis

Raymundo

et al. 2019

IFIP Advances in Information and Communication Technology

View full text Add to dashboard Cite

Competitiveness of Brazilian agribusiness depends on the capacity to sell products through the largest port in South America, Santos. The aim of this paper is to analyze the port efficiency indicators of Brazilian National Waterways Transportation Agency regarding the Port of Santos. To this end, we collected the indicator of port for handling, service time, and tariffs (fee) during of period of 2016-2018 and analyzed using time series. The results showed a tendency of slight growth for handling, moderate growth for the tariff, and a slight decrease in the service time.

show abstract

Time series feature learning with labeled and unlabeled data

Cited by 108 publications

References 14 publications

Self-supervised Learning for Semi-supervised Time Series Classification

Self-supervised Learning for Semi-supervised Time Series Classification

SelfMatch: Robust semisupervised time‐series classification with self‐distillation

Port Performance Measures in Brazil: An Analysis in Port of Santos

Contact Info

Product

Resources

About