Ensembles of Deep LSTM Learners for Activity Recognition using Wearables
Open accessThe DOI matches its Crossref record; 53 of the 65 references it cites were found in a registry or an index, and 10 are not checked yet; none of the 62 sources its claims cite is checked for support yet.
Abstract
Recently, deep learning (DL) methods have been introduced very successfully into human activity recognition (HAR) scenarios in ubiquitous and wearable computing. Especially the prospect of overcoming the need for manual feature design combined with superior classification capabilities render deep neural networks very attractive for real-life HAR applications. Even though DL-based approaches now outperform the state-of-the-art in a number of recognition tasks, still substantial challenges remain. Most prominently, issues with real-life datasets, typically including imbalanced datasets and problematic data quality, still limit the effectiveness of activity recognition using wearables. In this paper we tackle such challenges through Ensembles of deep Long Short Term Memory (LSTM) networks. LSTM networks currently represent the state-of-the-art with superior classification performance on relevant HAR benchmark datasets. We have developed modified training procedures for LSTM networks and combine sets of diverse LSTM learners into classifier collectives. We demonstrate that Ensembles of deep LSTM learners outperform individual LSTM networks and thus push the state-of-the-art in human activity recognition using wearables. Through an extensive experimental evaluation on three standard benchmarks (Opportunity, PAMAP2, Skoda) we demonstrate the excellent recognition capabilities of our approach and its potential for real-life applications of human activity recognition.
Checked 2026-10-05 · 10 references not checked yet
Overview
CoChat Summary
TL;DR
The paper introduces ensembles of deep LSTM learners for wearable activity recognition, using an epoch-wise bagging-style training procedure and score-level fusion of multiple LSTM models. Across Opportunity, PAMAP2, and Skoda, the ensembles outperform individual LSTMs and state-of-the-art baselines, with the largest gains on the challenging, imbalanced Opportunity dataset.
Key findings
- Ensembles of LSTM learners improve over single LSTMs on all three benchmarks. For Opportunity, single CE LSTM F1 = 0.674 ± 0.021 vs. Ensemble M=10 CE = 0.716 ± 0.011 and M=20 CE = 0.720 ± 0.011; the best combined CE+F1 ensemble reaches 0.726 ± 0.008. For PAMAP2, single CE = 0.802 ± 0.045 vs. Ensemble M=10 CE = 0.848 ± 0.021 and M=20 CE = 0.852 ± 0.026; combined CE+F1 = 0.854 ± 0.026. For Skoda, single CE = 0.918 ± 0.007 vs. Ensemble M=10 CE = 0.925 ± 0.002 and M=20 CE = 0.926 ± 0.001; combined CE+F1 = 0.924 ± 0.002.
- The LSTM ensemble outperforms reported state-of-the-art baselines: LSTM baseline 0.659/0.756/0.904, DeepConvLSTM 0.672/0.748/0.912, and LSTM-S 0.684/0.838/0.921 for Opportunity/PAMAP2/Skoda, respectively, versus the ensemble’s 0.726/0.854/0.924.
- Performance generally plateaus at M=10 base learners; increasing to M=20 gives no statistically significant improvement except for Opportunity with F1 loss, where the improvement is “just significant.” Combining CE and F1 loss functions further improves Opportunity but not PAMAP2, and Skoda shows no substantial boost.
- The training procedure intentionally omits data: on average about 35%–38% of training data is not used per epoch when using a uniformly distributed random number generator for training frames.
Methods
- The authors use two-layer LSTM networks with 256 LSTM units per layer, dropout 0.5 at both hidden layers, ADAM with learning rate 0.001, implemented in Lasagne/Theano and trained on an NVIDIA Tesla K40. Training uses randomly drawn mini-batch sizes from U(128, 256) and frame lengths from U(16, 32); test-time inference is sample-wise with mini-batch size and frame length set to 1.
- The core method is epoch-wise bagging: for each epoch, random mini-batch size, random starting positions, and random frame lengths are drawn; one LSTM learner is produced per epoch (MaxEpoch = 100). The M best models on a validation set are retained and fused by arithmetic mean of their softmax class-probability vectors. Both cross-entropy and F1 loss are explored.
- Evaluation uses three standard benchmarks: Opportunity, PAMAP2, and Skoda, with hold-out validation/test protocols replicating prior work. Results are mean F1 scores over 30 repetitions, with two-tailed independent t-tests for significance.
Limitations
- The paper states that implementation issues for field deployments or products are not yet addressed; ensembles inflate the parameter space, though the authors argue pruning remains accessible because the base LSTM models are unchanged.
- The paper notes that F1 loss is a global measure and cannot directly measure individual sample losses; because LSTMs are trained on mini-batches, it may not precisely approximate the true gradient toward optimal F1 over the whole training set.
- The paper states that Skoda shows no substantial performance boost, attributed to only one worker being considered and the task lacking challenging test data from different environments; the baseline models already achieve high specialisation (F1 > 0.9). Evident from the text: evaluation is limited to three benchmark datasets and does not include an in-the-wild deployment, and frame-based results from prior work were converted to sample-wise predictions for comparability.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 11 Not valid · 48 Not checked · 80 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Through an extensive experimental evaluation on three standard benchmarks (Opportunity, PAMAP2, Skoda) we demonstrate the excellent recognition capabilities of our approach and its potential for real-life applications of human activity recognition.”
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“By outperforming the state-of-the-art our results indicate increased robustness of our Ensembles of deep LSTM learners with regards to typical challenges of human activity recognition in ubiquitous and wearable computing scenarios.”
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We demonstrate that Ensembles of deep LSTM learners outperform individual LSTM networks and thus push the state-of-the-art in human activity recognition using wearables.”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“strong learners -LSTM models in our case- can indeed be integrated into very effective ensembles that outperform single, that is, individual models. Recognition results are consistent across all three tasks”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“PAMAP2 does not benefit from diversity in loss function”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
55 more claims
Create a free account to see all 139 claims and to verify the 62 references still open.
- It may state two separate claims as one. 32
- The quote only partly states this claim. 2
- Probably not a substantial claim. 59
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Many wearables datasets exhibit substantial class imbalance [68].”
From Past, Present, and Future of Sensor-based Human Activity Recognition Using Wearables: A Surveying Tutorial on a Still Challenging Task (2025) · cites this paper as [68]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 3.2 Modified Training for LSTM Models · p. 12
“Overall, only a total of ˆ B ⌊ T/ ˆ B ⌋ samples are used in each epoch.”
§ ALGORITHM 1: Epoch-wise bagging for LSTM learners. · p. 13
“while L seq ≤ ⌊ T/ ˆ B j ⌋ do draw random integer ˆ L i (i.e., frame length) from discrete uniform distribution U ( L low , L high ) ;”
Citations
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
The Elements of Statistical Learning
Hastie, T., Tibshirani, R., & Friedman, J. (2009) · Springer Series in Statistics · doi:10.1007/978-0-387-84858-7 · Cited by 25,970
The Elements of Statistical Learning
Hastie, T., Friedman, J., & Tibshirani, R. (2001) · Springer Series in Statistics · doi:10.1007/978-0-387-21606-5 · Cited by 24,162
Breiman, L. (1996) · Machine Learning · doi:10.1007/bf00058655 · Cited by 16,325
Parallel Distributed Processing: Explorations in the Microstructure of Cognition: Foundations
Rumelhart, D. E., McClelland, J. L., & AU (1986) · doi:10.7551/mitpress/5236.001.0001 · Cited by 15,686
The Computer for the 21st Century
Weiser, M. (1991) · Scientific American · doi:10.1038/scientificamerican0991-94 · Cited by 10,017
Large-Scale Machine Learning with Stochastic Gradient Descent
Bottou, L. (2010) · Proceedings of COMPSTAT'2010 · doi:10.1007/978-3-7908-2604-3_16 · Cited by 5,795
Research works printed but not linked: 2
- James L McClelland and David E Rumelhart . 1986. Parallel distributed processing . MIT Press . James L McClelland and David E Rumelhart. 1986. Parallel distributed processing. MIT Press.
- Theano Development Team . 2016 . Theano: A Python framework for fast computation of mathematical expressions. arXiv e-prints abs/1605.02688 (May 2016). http://arxiv.org/abs/1605.02688 Theano Development Team. 2016. Theano: A Python framework for fast computation of mathematical expressions. arXiv e-prints abs/1605.02688 (May 2016). http://arxiv.org/abs/1605.02688 · arXiv
Cited by
Items 201–300 of 539
Sensor-based Human Activity Recognition Using Graph LSTM and Multi-task Classification Model
Cao, J., Wang, Y., Tao, H., & Guo, X. (2022) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3561387 · Cited by 25
Zhu, Y., Luo, H., Chen, R., & Zhao, F. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3285547 · Cited by 25
“UESTC_IndRNN [18] adopted the independently RNN (IndRNN) on the spectrum to obtain short-and long-time features for human behavior recognition.”
MethodPrediction of freezing of gait based on self-supervised pretraining via contrastive learning
Xia, Y., Sun, H. D., Zhang, B., Xu, Y., & Ye, Q. (2023) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2023.105765 · Cited by 24
HARNet: Towards On-Device Incremental Learning using Deep Ensembles on Constrained Devices
Sundaramoorthy, P., Gudur, G. K., Moorthy, M. R., Bhandari, R. N., & Vijayaraghavan, V. (2018) · International Workshop on Embedded and Mobile Deep Learning (EMDL) · doi:10.1145/3212725.3212728 · Cited by 23
“To learn hierarchial features, Guan et al. [4] proposed an ensem-bleofLSTMlearnersforbuildingarobustrecognitionsystem.”
BackgroundA Comparative Research on Human Activity Recognition Using Deep Learning
Tüfek, N., & Özkaya, Ö. (2019) · Signal Processing and Communications Applications Conference (SIU) · doi:10.1109/siu.2019.8806395 · Cited by 23
“Giyilebilir cihazlar üzerindeki algılayıcıları kullanarak insan aktivitelerinin algılanması probleminin derin öğrenme ile çözülmesi esnasında sık kullanılan çözümlerden biri de uzun kısa süreli bellek (LSTM) yöntemidir [18].”
Heart Attack Detection in Colour Images Using Convolutional Neural Networks
Rojas-Albarracín, G., Chaves, M. Á., Fernández‐Caballero, A., & López, M. T. (2019) · Applied Sciences · doi:10.3390/app9235065 · Cited by 25
GAN-based Style Transformation to Improve Gesture-recognition Accuracy
Suzuki, N., Watanabe, Y., & Nakazawa, A. (2020) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3432199 · Cited by 24
“In addition, to encode long-term correlations in sensory signals, several studies proposed combining 2D convolution and RNN/long short-term memory in order to learn inter-channel and long-temporal relationships [11, 13, 18, 29].”
BackgroundTemporal Action Localization for Inertial-based Human Activity Recognition
Bock, M., Moeller, M., & Laerhoven, K. V. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3699770 · Cited by 29
“Caused by frequently occuring activity switches, the fragmentation ultimately leads to significantly lower mAP scores being produced by the inertial architectures as opposed to the TAL models.”
Olugbade, T., Newbold, J. W., Johnson, R., Volta, E., Alborno, P., Niewiadomski, R., Dillon, M., Volpe, G., & Bianchi‐Berthouze, N. (2020) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2020.2978069 · Cited by 23
“Findings in body movement analysis in the context of activity recognition suggest that the LSTM layer can also learn by integration across temporal and non-temporal dimensions [21][22].”
Estimation of Blood Alcohol Concentration From Smartphone Gait Data Using Neural Networks
Li, R., Balakrishnan, G. P., Nie, J., Li, Y., Agu, E., Grimone, K., Herman, D. S., Abrantes, A. M., & Stein, M. D. (2021) · IEEE Access · doi:10.1109/access.2021.3054515 · Cited by 24
“In the future, we plan on using an ensemble of bi-LSTM networks by extending [34].”
MethodTime to Think the Security of WiFi-Based Behavior Recognition Systems
Liu, J., He, Y., Xiao, C., Han, J., & Ren, K. (2023) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2023.3261328 · Cited by 28
“Traditional approaches utilize cameras [1], [2], [3], sonar[4],[5],[6],orwearabledevices[7],[8]tocapturebehavior information,includinggesture,activity,andthelike.”
MethodHaque, S. T., Debnath, M., Yasmin, A., Mahmud, T., & Ngu, A. H. H. (2024) · Sensors · doi:10.3390/s24196235 · Cited by 25
“As pointed out by [9], there are a few inherent challenges in applying deep learning to wearable devices such as smartwatches.”
BackgroundMV-Sports: A Motion and Vision Sensor Integration-Based Sports Analysis System
Zhang, C., Yang, F., Li, G., Zhai, Q., Jiang, Y. F., & Xuan, D. (2018) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom.2018.8485910 · Cited by 22
“[19] employ deep recurrent LSTM networks for action recognition using wearable sensing data.”
BackgroundPPG-based Heart Rate Estimation with Time-Frequency Spectra: A Deep Learning Approach
Reiss, A., Schmidt, P., Indlekofer, I., & Laerhoven, K. V. (2018) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3267305.3274176 · Cited by 26
“Therefore, since capturing more data variability during training, combining these models could lead to increased modelling capabilities [6].”
BackgroundSelective Ensemble Based on Extreme Learning Machine for Sensor-Based Human Activity Recognition
Tian, Y., Zhang, J., Chen, L., Geng, Y., & Wang, X. (2019) · Sensors · doi:10.3390/s19163468 · Cited by 22
Hierarchical Self Attention Based Autoencoder for Open-Set Human Activity Recognition
Tonmoy, M. T. H., Mahmud, S., Rahman, A. K. M. M., Amin, M. A., & Ali, A. A. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-75768-7_28 · Cited by 22
“In that regard, various forms of convolutional, recurrent and hybrid architectures such as [7,8,16] were proposed and demonstrated varying levels of success in recognition of the activities under consideration.”
BackgroundMeena, T., & Sarawadekar, K. P. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3301187 · Cited by 21
“An ensembled HAR model [29] based on the combination of deep LSTM was proposed which achieved a high accuracy on”
MethodChen, H., Xiong, F., Wu, D., Zheng, L., Peng, A., Hong, X., Tang, B., Lü, H., Shi, H., & Zheng, H. (2017) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm.2017.8217821 · Cited by 23
“Many reports are comparing the results of multiple database identification, but do not propose the difference between the volume of different database data [15], [16].”
Yu, H., Pan, G., Pan, M., Li, C., Jia, W., Zhang, L., & Sun, M. (2019) · Sensors · doi:10.3390/s19030546 · Cited by 24
Abedin, A., Rezatofighi, S. H., Shi, Q., & Ranasinghe, D. C. (2019) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2019/801 · Cited by 29
Nakandala, S., Jankowska, M. M., Tuz‐Zahra, F., Bellettiere, J., Carlson, J., LaCroix, A. Z., Hartman, S. J., Rosenberg, D. E., Zou, J., Kumar, A., et al. (2021) · Journal for the Measurement of Physical Behaviour · doi:10.1123/jmpb.2020-0016 · Cited by 22
“Two main types of deep models have been applied in activity recognition: convolutional neural networks (CNNs) (Krizhevsky, Sutskever, & Hinton, 2012) and long short-term memory models (Guan & Plötz, 2017).”
HAWAD: Hand Washing Detection using Wrist Wearable Inertial Sensors
Mondol, M. A. S., & Stankovic, J. A. (2020) · International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT) · doi:10.1109/dcoss49796.2020.00016 · Cited by 26
“[8] ensembles a set of Long Short Term Memory (LSTM) networks to improve the performance of activity recognition tasks.”
BackgroundMethodXia, S., Chu, L., Pei, L., Yu, W., & Qiu, R. C. (2022) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2022.3173957 · Cited by 20
“1) LSTM Baseline [44]: An recurrent neural network (RNN) network with multilayers LSTM for HAR based on inertial sensor time-series.”
BackgroundEnsemble Deep Learning on Wearables Using Small Datasets
Mauldin, T. R., Ngu, A. H. H., Metsis, V., & Canby, M. E. (2020) · ACM Transactions on Computing for Healthcare · doi:10.1145/3428666 · Cited by 22
“One way to overcome this real-world limitations is to combine sets of diverse LSTM learners into an ensemble of classifiers as shown in Reference [11].”
Continual Activity Recognition with Generative Adversarial Networks
Ye, J., Nakwijit, P., Schiemer, M., Jha, S., & Zambonelli, F. (2021) · ACM Transactions on Internet of Things · doi:10.1145/3440036 · Cited by 23
“A large number of modern data-driven techniques, including Hidden Markov Models, Support Vector Machines, and more recently deep neural networks [49], have achieved promising results in extracting features from raw sensor data [19, 29] and learning complex correlations between sensor data and activities of interest [1…”
BackgroundBlock-Wise Training Residual Networks on Multi-Channel Time Series for Human Activity Recognition
Teng, Q., Zhang, L., Tang, Y., Song, S., Wang, X., & He, J. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3085360 · Cited by 21
“We also compare the block-wise training residual ResNet and its two variants with the state-of-the-art methods, which are from those literatures such as [32] [26] [24] [33], accordingly (Table I).”
MethodTS2ACT: Few-Shot Human Activity Sensing with Cross-Modal Co-Learning
Xia, K., Li, W., Gan, S., & Lu, S. (2023) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3631445 · Cited by 19
On specialized window lengths and detector based human activity recognition
Li, H., Abowd, G. D., & Plötz, T. (2018) · Proceedings of the 2018 ACM International Symposium on Wearable Computers · doi:10.1145/3267242.3267246 · Cited by 19
“In contrast to the periodic activities covered by most of the other datasets, activities in Opportunity are largely sporadic and vary a lot (in duration, and general appearance) within classes [8, 19].”
BackgroundSelf-citationNon-invasive technique for real-time myocardial infarction detection using faster R-CNN
Mohan, H., Rao, P. V., Kumara, H. C. S., & Manasa, S. (2021) · Multimedia Tools and Applications · doi:10.1007/s11042-021-10957-2 · Cited by 20
“Wearable devices that use different kinds of sensors, like gyroscopes, micro-phones accelerometers, or combinations of all these can fall into the Non-visual approach category [2, 32, 40, 55].”
BackgroundArakawa, R., Ahuja, K., Mak, K. J., Thompson, G., Shaaban, S., Lindhiem, O. J., & Goel, M. (2023) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3596244 · Cited by 24
“Note that several existing approaches use neural networks for human activity recognition, such as using Convolutional Neural Network (CNN) [29] and Long Short-Term Memory Network (LSTM) [26].”
MethodMulti-modal egocentric activity recognition using multi-kernel learning
Arabacı, M. A., Özkan, F., Sürer, E., Jančovič, P., & Temizel, A. (2020) · Multimedia Tools and Applications · doi:10.1007/s11042-020-08789-7 · Cited by 22
“In the literature, sensor-based deep network solutions for HAR use generative (Restricted Boltzman Machines [44], Deep Autoencoders [45]), discriminative (Convolutional Neural Networks [46], Recurrent Neural Networks [47]) or hybrid (Convolutional Recurrent Neural Networks [48]) models.”
MethodChronic Pain Protective Behavior Detection with Deep Learning
Wang, C., Olugbade, T., Mathur, A., Williams, A. C. D. C., Lane, N. D., & Bianchi‐Berthouze, N. (2021) · ACM Transactions on Computing for Healthcare · doi:10.1145/3449068 · Cited by 20
“On the other hand, a sample-wise prediction is also conducted following [36], where each output state Ht would be used as input for a fully-connected layer with softmax activation for classification instead of just using the last output HT .”
BackgroundMethodInfluentialContext-Aware Complex Human Activity Recognition Using Hybrid Deep Learning Models
Omolaja, A., Otebolaku, A., & Alfoudi, A. S. D. (2022) · Applied Sciences · doi:10.3390/app12189305 · Cited by 21
“They consume a lot of computational resources and time [46–49].”
BackgroundMethodImproving Cross-Subject Activity Recognition via Adversarial Learning
Leite, C. F. S., & Xiao, Y. (2020) · IEEE Access · doi:10.1109/access.2020.2993818 · Cited by 18
“The performances for DeepConvLSTM [26] and the LSTM with Uniqueness Attention [41] are also reported.”
BackgroundMethodInfluentialYadav, R. K., & Pradhan, A. K. (2020) · IEEE Systems Journal · doi:10.1109/jsyst.2020.3012937 · Cited by 18
Shakerian, R., Yadollahzadeh‐Tabari, M., & Rad, S. Y. B. (2022) · IET Biometrics · doi:10.1049/bme2.12066 · Cited by 21
HarMI: Human Activity Recognition Via Multi-Modality Incremental Learning
Zhang, X., Yu, H., Yang, Y., Gu, J., Li, Y., Zhuang, F., Yu, D., & Ren, Z. (2021) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2021.3085602 · Cited by 24
A hybrid CNN and BLSTM network for human complex activity recognition with multi-feature fusion
Huan, R., Zhan, Z., Ge, L., Chi, K., Chen, P., & Liang, R. (2021) · Multimedia Tools and Applications · doi:10.1007/s11042-021-11363-4 · Cited by 18
Toward Robust Device-Free Gesture Recognition Based on Intrinsic Spectrogram of mmWave Signals
Wu, J., Wang, J., Gao, Q., Cheng, M., Pan, M., & Zhang, H. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3165196 · Cited by 20
A Spatiotemporal Ensemble Approach to Rainfall Forecasting
Souto, Y. M., Porto, F. A. M., Moura, A. M., & Bezerra, E. A. D. (2018) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2018.8489693 · Cited by 20
“Other interesting works define ensemble methods using deep neural networks architectures [15][22] [23][24][25][26][27].”
BackgroundFilterNet: A Many-to-Many Deep Learning Architecture for Time Series Classification
(2020) · doi:10.3390/s20092498 · Cited by 17
“It is likely that an ensembling strategy such as that used in [39], combined with the pooling and multiscale features exhibited in this work, could improve performance even further.”
BackgroundMethodInfluentialA Human Activity Recognition Model Based on Wearable Sensor
Zhou, F., Wang, R., Su, H., & Xu, S. (2022) · International Conference on Digital Home (ICDH) · doi:10.1109/icdh57206.2022.00033 · Cited by 16
Teng, Q., Tang, Y., & Hu, G. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3364187 · Cited by 18
“Based on ensemble learning, [35] integrated multiple LSTM base learners into classifier groups and developed a modified training procedure to minimize empirical error.”
BackgroundXia, S., Chu, L., Pei, L., Yang, J., Yu, W., & Qiu, R. C. (2024) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2024.3381171 · Cited by 23
ProCAVIAR: Hybrid Data-Driven and Probabilistic Knowledge-Based Activity Recognition
Bettini, C., Civitarese, G., Giancane, D., & Presotto, R. (2020) · IEEE Access · doi:10.1109/access.2020.3015091 · Cited by 17
“driven component of ProCAVIAR with a state-of-the-art method based on deep learning [38].”
MethodSOFA-Net: Second-Order and First-order Attention Network for Crowd Counting
Duan, H., Wang, S., & Guan, Y. (2020) · British Machine Vision Conference (BMVC) · doi:10.5244/c.34.62 · Cited by 16
Self-citationYONO: Modeling Multiple Heterogeneous Neural Networks on Microcontrollers
Kwon, Y. D., Chauhan, J., & Mascolo, C. (2022) · ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1109/ipsn54338.2022.00030 · Cited by 20
Primitive activity recognition from short sequences of sensory data
Mohammad, Y. F. O., Matsumoto, K., & Hoashi, K. (2018) · Applied Intelligence · doi:10.1007/s10489-018-1166-6 · Cited by 16
“PAR systems can be further divided into systems that deal with the problem as a temporal time-series classification task employing approaches like Hidden Markov Models (HMM) [28, 32, 36] or Long-Short Term Memory (LSTM) [15, 16, 26], and systems that treat it as a standard classification problem from short sequences o…”
BackgroundInfluentialDeep feature learning and selection for activity recognition
Mohammad, Y. F. O., Matsumoto, K., & Hoashi, K. (2018) · ACM Symposium on Applied Computing · doi:10.1145/3167132.3167234 · Cited by 21
“the problem as a temporal time-series classification task employing approaches like HiddenMarkovModels (HMM) [28, 30, 31] or LongShort TermMemory (LSTM) [13, 14, 27], and systems that treat it as a standard classification problem from short sequences of sensory”
MethodInfluentialLiu, X., Cheng, D., Fu, Z., Zhang, L., Wu, H., & Song, A. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3664424 · Cited by 16
“Plötz et al. [25] proposed to ensemble multiple deep LSTM learners, which considerably outperform an individual LSTM in activity recognition tasks using wearables.”
ContrastSense: Domain-invariant Contrastive Learning for In-the-Wild Wearable Sensing
Dai, G., Xu, H., Yoon, H., Li, M., Tan, R., & Lee, S.-J. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3699744 · Cited by 27
“To unleash the full potential of wearable sensing data, deep learning techniques have been widely adopted, leading to notable advancements in performance [13, 30, 67].”
Qian, B., Su, J., Wen, Z., Jha, D. N., Li, Y., Guan, Y., Puthal, D., James, P., Yang, R., Zomaya, A. Y., et al. (2019) · White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · doi:10.48550/arxiv.1910.05433 · Cited by 16
“the hidden units for "better memories", came into popularity in various applications such as speech recognition [95], video analysis [265], language translation [161], activity recognition [97]. etc. Since data streaming the most common form in the IoT environment, RNN(LSTM) is deemed as one of the most powe…”
MethodSelf-citationDeep Learning for Detecting Human Activities From Piezoelectric-Based Kinetic Energy Signals
Manjarrés, J., Lan, G., Gorlatova, M. A., Hassan, M., & Pardo, M. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3093245 · Cited by 16
Lakoju, M., Ajienka, N., Khanesar, M. A., Burnap, P., & Branson, D. T. (2021) · Sensors · doi:10.3390/s21154991 · Cited by 16
Khatun, M. A., Yousuf, M., & Moni, M. A. (2023) · International Conference on Electrical, Computer and Communication Engineering · doi:10.1109/ecce57851.2023.10101550 · Cited by 19
“For the opportunity dataset, it also outperformed the most recent technique, which combines CNNS and unidirectional RNNs [11].”
BackgroundMethodTahir, S. B. U. D., Jalal, A., & Kim, K. (2021) · International Bhurban Conference on Applied Sciences and Technology (IBCAST) · doi:10.1109/ibcast51254.2021.9393204 · Cited by 15
Haresamudram, H., Essa, I., & Plötz, T. (2024) · Sensors · doi:10.3390/s24041238 · Cited by 14
“In recent years, the use of supervised deep learning [10,11] has resulted in great improvements in performance, relative to more traditional methods involving heuristics [14].”
Self-citationKwon, H., Tong, C., Haresamudram, H., Gao, Y., Abowd, G. D., Lane, N. D., & Ploetz, T. (2020) · arXiv · doi:10.48550/arxiv.2006.05675 · Cited by 14
“available datasets for human activity recognition contain imbalanced classes. For example, approximately 75% of the Opportunity dataset (which has 18 classes in total) [14] consists of the null class [27], making it challenging to design classifiers. The activities being studied also impact the class imbalance to some…”
BackgroundInfluentialNisar, M. A., Shirahama, K., Irshad, M. T., Huang, X., & Grzegorzek, M. (2023) · Sensors · doi:10.3390/s23198234 · Cited by 17
“Guan et al. [30] presents LSTM networks for HAR using time-series data.”
MethodDevice-Free Wireless Sensing With Few Labels Through Mutual Information Maximization
Chen, B., Wang, J., Lv, Y., Gao, Q., Pan, M., & Fang, Y. M. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3325948 · Cited by 14
“Most of the traditional sensing methods use wearable devices [14], [15], [16] and vision-based systems [17], [18].”
BackgroundHardenVR: Harassment Detection in Social Virtual Reality
Wang, N., Zhou, J., Li, J., Han, B., Li, F., & Chen, S. (2024) · IEEE Conference on Virtual Reality and 3D User Interfaces (VR) · doi:10.1109/vr58804.2024.00033 · Cited by 15
“There is plenty of research on human activity recognition (HAR) [11,25,44,64,65,67].”
BackgroundMethodAn Implementation of Hybrid CNN-LSTM Model for Human Activity Recognition
Krishna, K. S., & Paneerselvam, S. (2022) · Lecture notes in electrical engineering · doi:10.1007/978-981-19-1111-8_63 · Cited by 14
Multi-agent Transformer Networks for Multimodal Human Activity Recognition
Li, J., Yao, L., Li, B., Wang, X., & Sammut, C. (2022) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3511808.3557402 · Cited by 15
“Recent approaches utilized deep neural networks, such as CNN [2, 10, 26, 41, 50, 51, 54] and LSTM Networks [5, 18, 35, 38, 53], as feature extractors to learn the representation of the input sensory segments automatically, then map the representation to labels using another neural network [3].”
MethodA Light-Weight Artificial Neural Network for Recognition of Activities of Daily Living
Mohamed, S. A., & Martínez-Hernández, U. (2023) · Sensors · doi:10.3390/s23135854 · Cited by 14
“The third benchmarking approach used an LSTM network following the approach proposed in [47,48].”
MethodPhase Randomization: A data augmentation for domain adaptation in human action recognition
Mitsuzumi, Y., Irie, G., Kimura, A., & Nakazawa, A. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110051 · Cited by 14
A dataset for activity recognition in an unmodified kitchen using smart-watch accelerometers
Mohammad, Y. F. O., Matsumoto, K., & Hoashi, K. (2017) · International Conference on Mobile and Ubiquitous Multimedia (MUM) · doi:10.1145/3152832.3152844 · Cited by 17
“A conceptually similar approach that showed promise in different activity recognition tasks is the use recurrent methods like LSTM in an online setting [15, 16].”
MethodPredicting progression patterns of type 2 diabetes using multi-sensor measurements
Ramazi, R., Perndorfer, C., Soriano, E. C., Laurenceau, J., & Beheshti, R. (2021) · Smart Health · doi:10.1016/j.smhl.2021.100206 · Cited by 17
Designing compact features for remote stroke rehabilitation monitoring using wearable accelerometers
Chen, X., Guan, Y., Shi, J. Q., Du, X., & Eyre, J. A. (2023) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-022-00124-3 · Cited by 16
“,[11][30]), the annotation here is inadequate.”
BackgroundSelf-citationChen, X., Zhou, X., Sun, M., & Wang, H. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3491933 · Cited by 13
“With the rapid development of deep learning techniques, many supervised deep learning models have been successfully applied to HAR [1], [2], [3].”
Lee, E.-J., Kim, K., & Park, J. (2022) · Frontiers in Marine Science · doi:10.3389/fmars.2022.1037697 · Cited by 13
“In machine learning, the ensemble of models makes the model performance more robust (Opitz and Maclin, 1999), which has been demonstrated in its application to LSTM (Guan and Plötz, 2017).”
MethodInfluentialUhlenberg, L., Haeusler, L. O., & Amft, O. (2024) · IEEE Access · doi:10.1109/access.2024.3513477 · Cited by 16
“Additionally, training parameters were determined through an automated search on participant P5 as validation participant in the tuning process, while participant P6 was excluded from the search, according to the validation and tests set definition in literatures [40] and [61].”
Guo, X., Kim, Y., Ning, X., & Min, S. D. (2025) · Sensors · doi:10.3390/s25020301 · Cited by 14
Enhancing Representation of Deep Features for Sensor-Based Activity Recognition
Li, X., Nie, L., Si, X., Ding, R., & Zhan, D. (2020) · Mobile Networks and Applications · doi:10.1007/s11036-020-01689-y · Cited by 13
“Therefore, deep learning approaches, including the widely used restricted Boltzmann machines [26], CNN [27, 28], recurrent neural networks (RNN) [29], LSTM [30] and their composition models [31, 32] have been applied and proven to be superior to the state-of-the-art methods in a number of recognition tasks.”
MethodInfluentialApplying Deep Learning Techniques to Estimate Patterns of Musical Gesture
Dalmazzo, D., Waddell, G., & Ramírez, R. (2021) · Frontiers in Psychology · doi:10.3389/fpsyg.2020.575971 · Cited by 17
“Guan and Plötz (2017) applied an LSTM architecture fed with IMU data taken from the same “Opportunity” activity recognition dataset.”
MethodHAR-GCNN: Deep Graph CNNs for Human Activity Recognition From Highly Unlabeled Mobile Sensor Data
Mohamed, A., Lejarza, F., Cahail, S., Claudel, C., & Thomaz, E. (2022) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops53856.2022.9767342 · Cited by 21
“[20] reported improved HAR using ensembles of deep LSTM learners relative to previously reported recurrent neural networks.”
BackgroundDevelompent of Activity Recognition Model using LSTM-RNN Deep Learning Algorithm
Gaur, D., & Dubey, S. K. (2022) · Journal of information and organizational sciences · doi:10.31341/jios.46.2.1 · Cited by 13
Personalized Fall Detection System
Ngu, A. H. H., Metsis, V., Coyne, S., Chung, B., Pai, R., & Chang, J. (2020) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops48775.2020.9156172 · Cited by 17
“One way to overcome this real life limitations is to combine sets of diverse LSTM learners into an ensemble of classifiers as shown in [14].”
Bi-STAN: bilinear spatial-temporal attention network for wearable human activity recognition
Gao, C., Chen, Y., Jiang, X., Hu, L., Zhao, Z., & Zhang, Y. (2023) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-023-01781-1 · Cited by 13
“In recent years, the supervised DNN models, such as CNN [11], RNN [12] and Autoencoders [13], usually outperform traditional machine learning-based techniques.”
MethodKumar, P., & Chauhan, S. (2021) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-021-00063-5 · Cited by 13
RetractedMulti-Sensor Time-Series Classification for Activity Tracking Under Variable Length
Norgaard, S., Saeedi, R., & Gebremedhin, A. H. (2019) · IEEE Sensors Journal · doi:10.1109/jsen.2019.2953938 · Cited by 13
Prototype Similarity Learning for Activity Recognition
Bai, L., Yao, L., Wang, X., Kanhere, S. S., & Xiao, Y. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-47426-3_50 · Cited by 14
Chen, Z., Zhang, X., & Cheng, X. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i6.20584 · Cited by 13
Bulugu, I. (2023) · Signal Image and Video Processing · doi:10.1007/s11760-023-02545-8 · Cited by 13
“Traditional gesture recognition meth-ods mainly rely on pre-installed cameras [1–3] or use of wearable sensors [4–6], and use of some special equipment to realize contactless gesture recognition, such as voice sensors [7] and environmental detection sensors [8, 9].”
BackgroundEfficient Human Gait Activity Recognition Based on Sensor Fusion and Intelligent Stacking Framework
Tarekegn, A. N., Sajjad, M., Cheikh, F. A., Ullah, M., & Muhammad, K. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3319353 · Cited by 12
“For instance, Guan and Plötz [31] proposed ensemble techniques based on DL for small datasets to analyze time series data.”
BackgroundMethodMLPs Are All You Need for Human Activity Recognition
Ojiako, K., & Farrahi, K. (2023) · Applied Sciences · doi:10.3390/app132011154 · Cited by 12
Distilled mid-fusion transformer networks for multi-modal human activity recognition
Li, J., Yao, L., Li, B., & Sammut, C. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114160 · Cited by 13
Human-Aware Collaborative Robots in the Wild: Coping with Uncertainty in Activity Recognition
Yalçinkaya, B., Couceiro, M. S., Soares, S., & Valente, A. L. G. (2023) · Sensors · doi:10.3390/s23073388 · Cited by 12
“On the other hand, Guan and Plötz have combined multiple LSTM networks in an ensemble approach and obtained superior results [45].”
ResultDeriving Effective Human Activity Recognition Systems through Objective Task Complexity Assessment
Hiremath, S. K., & Plötz, T. (2020) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3432227 · Cited by 11
“, by leveraging ensembles of deep learners [12].”
BackgroundSelf-citationGhorrati, Z., Esmaeili, A., & Matson, T. E. (2024) · IEEE Access · doi:10.1109/access.2024.3462515 · Cited by 12
Automated Pipeline for Robust Cat Activity Detection Based on Deep Learning and Wearable Sensor Data
Mozumder, M. A. I., Armand, T. P. T., Sumon, R. I., Uddin, S. M. I., & Kim, H.-C. (2024) · Sensors · doi:10.3390/s24237436 · Cited by 11
RAG-HAR: Retrieval Augmented Generation-based Human Activity Recognition
Sivaroopan, N., Karunarathna, H., Madarasingha, C., Jayasumana, A. P., & Thilakarathna, K. (2026) · IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom67906.2026.11524560 · Cited by 15
“RNNs, were then adopted to better capture short and long-term relationships [14].”
EspialCog: General, Efficient and Robust Mobile User Implicit Authentication in Noisy Environment
Zhu, T., Weng, Z., Song, Q., Yuan, C., Liu, Q., Yan, C., Lv, M., & Chen, T. (2020) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2020.3012491 · Cited by 11
“Few works used LSTM for motion sensor-based user authentication, others such as [28, 29] utilized LSTM for wearable activity recognition, but the noisy labels in the complex environment and a large amount of real-world data were not considered by them.”
Yang, G., Zhang, L., Bu, C., Wang, S., Wu, H., & Song, A. (2023) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2023.3321639 · Cited by 14
“Plötz et al. [18] develop a framework through an ensemble classifier of multiple deep networks for activity recognition using sensor data.”
BackgroundNovel Human Activity Recognition by graph engineered ensemble deep learning model
Ghalan, M., & Aggarwal, R. K. (2024) · IFAC Journal of Systems and Control · doi:10.1016/j.ifacsc.2024.100253 · Cited by 12
TSFS: An Integrated Approach for Event Segmentation and ADL Detection in IoT Enabled Smarthomes
Mallick, M., Kodeswaran, P., Sen, S., Kokku, R., & Ganguly, N. (2018) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2018.2880206 · Cited by 11
“Similar to the domain of infrastructural sensors, there have been work on active learning [37], also several deep learning models have been recently proposed [38].”
BackgroundOn-Device Lumbar-Pelvic Movement Detection Using Dual-IMU: A DNN-Based Approach
Zhang, Y., Haghighi, P. D., Burstein, F. V., Yao, L., & Cicuttini, F. (2021) · IEEE Access · doi:10.1109/access.2021.3074755 · Cited by 11
“…our proposed CNN-LSTM model with four traditional machine learning models including (kNN, CART, SVM and Naïve Bayes) and four LSTM-based DNN models including (standard CNN-LSTM with early fusion [24], [27], standard CNN-LSTM with sensor-based late fusion [24], [27], LSTM [25] and Bidir-LSTM [28]).”
BackgroundMethodImproving Wearable-Based Activity Recognition Using Image Representations
Guinea, A. S., Sarabchian, M., & Mühlhäuser, M. (2022) · Sensors · doi:10.3390/s22051840 · Cited by 12
“Details F 1 -Score Guan and Plötz [46] Ensembles of deep LSTM networks 0.9260 AttnSense [47] Attention with CNN and a GRU network 0.9310 Zeng et al. [36] LSTM + Continuous Temporal Attention 0.9381 DeepConvLSTM [7] Convolutional and LSTM network 0.9580 Our approach Image Representation of Time…”
MethodInfluentialDynamic Multi-scale CNN Forest Learning for Automatic Cervical Cancer Segmentation
Bnouni, N., Rekik, I., Rhim, M. S., & Amara, N. E. B. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-00919-9_3 · Cited by 11
“All previous ensemble DL methods [7-16] developed for image segmentation and other tasks overlooked three major aspects, which if leveraged would further improve the target results: (1) the heterogeneity of MRI training patches distribution, (2) relational learning at multiple levels of CNN-to-CNN interactions in the…”
BackgroundMethodAtLAS: An Activity-Based Indoor Localization and Semantic Labeling Mechanism for Residences
Niu, X., Xie, L., Wang, J., Chen, H., Liu, D., & Chen, R. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3004496 · Cited by 14
Order Matters: Shuffling Sequence Generation for Video PredictionPreprint
Wang, J., Hu, B., Long, Y., & Guan, Y. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1907.08845 · Cited by 13
“There are recent variations of related work on human activity prediction and recognition [9, 11, 16, 17, 25], motion trajectory forecasting [1, 21], future frame prediction [24, 26] and so on.”
BackgroundSelf-citation
Metadata
- Authors
- Guan, Y., & Plötz, T.
- Affiliations
- Newcastle University (GB)
- Georgia Institute of Technology (US)
- ORCID
- Guan, Yu
- Venue
- Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Published
- 2017-06-30
- Volume
- 1
- Issue
- 2
- Pages
- 1-28
- ISSN
- 2474-9567 (ISSN-L)
- OpenAlex
- W2604630936
- Cited by
- 511
- References
- 65
- Citations by year
- 2026: 33 · 2025: 35 · 2024: 59 · 2023: 60 · 2022: 70 · 2021: 82 · 2020: 87 · 2019: 46 · 2018: 33 · 2017: 6
- FWCI
- 19.53
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Engineering and Physical Sciences Research Council (EP/M020576/1)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Computer Science › Human-Computer Interaction › Hand Gesture Recognition Systems
- Physical Sciences › Engineering › Biomedical Engineering › Gait Recognition and Analysis
- Keywords
- human activity recognition, wearable sensors, deep learning, ensemble classifiers, class imbalance, data quality, training protocols, PAMAP2, Opportunity dataset
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.