Curriculum learning
The DOI matches its Crossref record; 35 of the 40 references it cites were found in a registry or an index.
Abstract
Humans and animals learn much better when the examples are not randomly presented but organized in a meaningful order which illustrates gradually more concepts, and gradually more complex ones. Here, we formalize such training strategies in the context of machine learning, and call them "curriculum learning". In the context of recent research studying the difficulty of training in the presence of non-convex training criteria (for deep deterministic and stochastic neural networks), we explore curriculum learning in various set-ups. The experiments show that significant improvements in generalization can be achieved. We hypothesize that curriculum learning has both an effect on the speed of convergence of the training process to a minimum and, in the case of non-convex criteria, on the quality of the local minima obtained: curriculum learning can be seen as a particular form of continuation method (a general strategy for global optimization of non-convex functions).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“Moreover, the lack of acoustical signals could even account for spatial disorientation 91,92.”
From The impact of multisensory integration and perceptual load in virtual reality settings on performance, workload and presence (2021) · cites this paper as [92]
In the citing paper
“This suggests that curriculum-style learning (Bengio et al., 2009), where simpler or archetypal problems are presented first to build a repository of valid heuristics, may potentially bootstrap performance.”
From Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory (2025) · cites this paper as Bengio et al., 2009
Citations
Cites
Reducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
A fast learning algorithm for deep belief nets
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006) · Neural Computation · doi:10.1162/neco.2006.18.7.1527 · Cited by 16,336
Learning Deep Architectures for AI
Bengio, Y. (2009) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000006 · Cited by 9,239
Self-citationExtracting and composing robust features with denoising autoencoders
Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P.-A. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390294 · Cited by 7,436
Self-citationA unified architecture for natural language processing: deep neural networks with multitask learning
Collobert, R., & Weston, J. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390177 · Cited by 5,058 · cited 6×
“In Collobert and Weston (2008), the main objective is to learn an embedding for the words as a side-e.ect of learning to compute this score.”
BackgroundMethodInfluentialSelf-citationLearning Deep Architectures for AI
Bengio, Y. (2009) · now publishers, Inc. eBooks · doi:10.1561/9781601982957 · Cited by 4,941
Self-citationGreedy layer-wise training of deep networks
Bengio, Y., Lamblin, P., Popovici, D., & Larochelle, H. (2007) · Advances in Neural Information Processing Systems 19 · doi:10.7551/mitpress/7503.003.0024 · Cited by 4,841
Self-citationA Neural Probabilistic Language Model
Bengio, Y., Ducharme, R., Vincent, P., Jauvin, C., Jaz, K., Hofmann, T., Poggio, T., & Shawe-Taylor, J. S. (2000) · Cited by 3,479
Self-citationLearning and development in neural networks: the importance of starting small
Elman, J. L. (1993) · Cognition · doi:10.1016/0010-0277(93)90058-4 · Cited by 1,988 · cited 8×
“The idea of training a learning machine with a curriculum can be traced back at least to Elman (1993).”
BackgroundMethodInfluentialRestricted Boltzmann machines for collaborative filtering
Salakhutdinov, R., Mnih, A., & Hinton, G. (2007) · Proceedings of the 24th international conference on Machine learning · doi:10.1145/1273496.1273596 · Cited by 1,885
Research works printed but not linked: 3
- 10.5555/89603 · doi:10.5555/89603
- 10.5555/866740 · doi:10.5555/866740
- 10.5555/525571 · doi:10.5555/525571
Cited by
Items 2,201–2,300 of 5,240
Expert Training: Task Hardness Aware Meta-Learning for Few-Shot ClassificationPreprint
Zhou, Y., Wang, Y., Cai, J., Yu, Z., Hu, Q., & Wang, W. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.06240 · Cited by 11
“Humans learn new tasks following an efficient easy-to-hard way, which has already inspired curriculum learning to make the classifier converge faster and perform better [15], [16].”
BackgroundMulti-label Contrastive Predictive CodingPreprint
Song, J., & Ermon, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.09852 · Cited by 41
“To this end, we use a curriculum learning [5] approach where we select α values from high to low throughout training: higher α has higher bias, lower variance and easier to learn, whereas lower α has lower bias, higher variance and harder to learn.”
MethodLearning Adaptive Embedding Considering Incremental ClassPreprint
Yang, Y., Sun, Z., Zhu, H., et al. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.13351 · Cited by 9
“To solve this problem, we employ a learnable curriculum clustering operator, which aims to conduct clustering from easy (distinguishable) to difficult (confused) instances via curriculum learning [41], [42], [43].”
MethodAdvSim: Generating Safety-Critical Scenarios for Self-Driving VehiclesPreprint
Wang, J., Pun, A., Tu, J., Manivasagam, S., Sadat, A., Casas, S., Ren, M., & Urtasun, R. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2101.06549 · Cited by 8
“First, we propose a curriculum learning (CL) [5] baseline where we first train on standard examples till convergence (easy examples), and then train on real challenging scenarios selected based on reachable actors (Sec 4.”
BackgroundST3D: Self-training for Unsupervised Domain Adaptation on 3D Object DetectionPreprint
Yang, J., Shi, S., Wang, Z., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.05346 · Cited by 9
“[39, 9, 8] employ curriculum learning [2] and separate cases by their difficulties to realize local sample-level curriculum.”
BackgroundMulti-task curriculum learning in a complex, visual, hard-exploration domain: Minecraft
Kanitscheider, I., Huizinga, J., Farhi, D., Guss, W. H., Houghton, B., Sampedro, R., Zhokhov, P., Baker, B., Ecoffet, A., Tang, J., et al. (2021) · arXiv · doi:10.48550/arxiv.2106.14876 · Cited by 9
“Automated curriculum learning There is an extensive literature on training neural networks with a curriculum [9], see [10] for an overview.”
BackgroundLanguage Models With Meta-information
Shi, Y. (2014) · Research Repository (Delft University of Technology) · doi:10.4233/uuid:d9a0ae1d-3336-4e43-bc3d-7a3a06461f54 · Cited by 22
“[17] opens some intriguing vistas of inquiry.”
BackgroundInfluentialBigger is not always better: The importance of human-scale language modeling for psycholinguistics
Wilcox, E. G., Hu, M. Y., Mueller, A., Warstadt, A., Choshen, L., Zhuang, C., Williams, A., Cotterell, R., & Linzen, T. (2025) · Journal of Memory and Language · doi:10.1016/j.jml.2025.104650 · Cited by 9
Estimating Labeling Quality with Deep Object Detectors
Haase-Schütz, C., Hertlein, H., & Wiesbeck, W. (2019) · 2019 IEEE Intelligent Vehicles Symposium (IV) · doi:10.1109/ivs.2019.8814144 · Cited by 8
“In Curriculum learning [21] training examples are ordered in a meaningful way to improve training progress.”
BackgroundCost-Sensitive Graph Convolutional Network With Self-Paced Learning for Hit-and-Run Analysis
Wan, J., & Zhu, S. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3314402 · Cited by 8
“In this paper, in order to guarantee the correctness of inferred pseudo-labels for subsequent training of GCN, the self-paced curriculum learning strategy [33], [34] is adopted to conduct semi-supervised learning with an ‘easy-to-hard’ scheme.”
BackgroundMethodCOCL: An Intelligent Framework for Enhancing Deep Learning-Based Vulnerability Detection
Li, W., Dou, S., Wu, Y., Li, C., & Liu, Y. (2023) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2023.3329670 · Cited by 7
“In 2009, Bengio et al. [23] officially name this training method as CL and prove that curriculum learning improves both the study effectiveness and the convergence speed of the model.”
MethodNeuroEscape: Ordered Escape Routing via Monte-Carlo Tree Search and Neural Network
Chen, Z., Ho, T.-Y., Schlichtmann, U., Chen, D., Liu, M., Yao, H., & Yin, X. (2023) · doi:10.1109/iccad57390.2023.10323718 · Cited by 10
“To enhance training efficiency, we adopt curriculum learning [23], in which the training process is performed on a sequence of data with increasing difficulty.”
MethodREQA: Coarse-to-fine assessment of image quality to alleviate the range effect
Li, B., & Huo, F. (2024) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2023.104043 · Cited by 12
Load Identification Based on Attention Semisupervised Curriculum Label Learning With AVME-HT Feature
Jiang, J., Wang, Z., Zhao, H., et al. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3380587 · Cited by 8
“This process is reiterated until all unlabeled data are pseudolabeled or the maximum iteration limit is attained. randomly [32], [33].”
BackgroundMachine Learning and Information Theory Concepts towards an AI Mathematician
Bengio, Y., & Malkin, N. (2024) · Bulletin of the American Mathematical Society · doi:10.1090/bull/1839 · Cited by 33
“Active learning is related to another interesting notion from machine learning that relates to mathematics education: curriculum learning [Bengio et al., 2009].”
Self-citationFinding core labels for maximizing generalization of graph neural networks
Fu, S., Ma, X., Zhan, Y., You, F., Peng, Q., Liu, T., Bailey, J. P., & Mandic, D. P. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106635 · Cited by 8
Sulpice, L., Owaki, D., & Hayashibe, M. (2024) · Advanced Robotics · doi:10.1080/01691864.2024.2442718 · Cited by 8
“In the training process, inspired by curriculum learning [34] which consist in changing the task over time, the target velocity increases every 500 epochs by 0.25 m/s, from 1 m/s to 3.5 m/s (except for Walk model that stop at 1.75 m/s).”
InfluentialDSTF: A Diversified Spatio-Temporal Feature Extraction Model for traffic flow prediction
Wang, X., Wang, X., Huang, F., Zou, F., Liao, L., & Zeng, R. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.129280 · Cited by 12
Solving Multiobjective Combinatorial Optimization via Learning to Improve Method
Ye, T., Zhang, Z., Zhang, Q., Chen, J., & Wang, J. (2025) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2025.3540424 · Cited by 13
“The curriculum learning (CL) strategy [34] is used to derive the initial state for better sample efficiency (lines 6–7).”
SST: Self-training with self-adaptive thresholding for semi-supervised learning
Zhao, S., Huang, H., Li, X., Chen, X., & Wang, R. (2025) · Information Processing & Management · doi:10.1016/j.ipm.2025.104158 · Cited by 11
“ST++ [84] extends traditional self-training methods by exploiting progressively refined pseudo-labels in a curriculum learning manner [85].”
A data-guided curriculum towards low-resource neural machine translation
Wang, J., Yang, L., Wang, J., Guan, Y., Bai, L., & Luo, H. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127673 · Cited by 11
Domain Adaptation for Vehicle Detection Under Adverse Weather
Lin, H., Huang, Y., Lai, J., & You, T.-T. (2025) · IEEE Open Journal of Intelligent Transportation Systems · doi:10.1109/ojits.2025.3563373 · Cited by 17
“In this work, we employ the concept of curriculum learning [33] to prioritize easier images for training before tackling more difficult samples.”
Curriculum Direct Preference Optimization for Diffusion and Consistency Models
Croitoru, F.-A., Hondru, V., Ionescu, R. T., Sebe, N., & Shah, M. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.00269 · Cited by 51
“In this work, we propose a novel and enhanced version of Direct Preference Optimization based on curriculum learning [6, 69], which is termed Curriculum DPO , to train text-to-image generative (diffusion [51] or consistency [63, 66]) models.”
InfluentialSpecialized curricula for training vision language models in retinal image analysis
Holland, R., Taylor, T. R. P., Holmes, C., Riedl, S., Mai, J., Patsiamanidi, M., Mitsopoulou, D., Hager, P., Müller, P., Paetzold, J. C., et al. (2025) · npj Digital Medicine · doi:10.1038/s41746-025-01893-8 · Cited by 26
“To train VLMs in these skills, we develop a curriculum-based approach 15 that draws from recent advances in instruction fi netuning 16 – 18 which iteratively re fi ne models on datasets of increasing quality.”
Improved PPO Optimization for Robotic Arm Grasping Trajectory Planning and Real-Robot Migration
Li, C., Liu, Z., Liang, L., Ji, Z., Li, C., Liang, J., & Li, Y. (2025) · Sensors · doi:10.3390/s25175253 · Cited by 11
“For example, the curriculum learning [27] framework adopts an incremental training paradigm to systematically increase task difficulty.”
PyraNet: A Multi-Layered Hierarchical Dataset for Verilog
Nadimi, B., Boutaib, G. O., & Zheng, H. (2025) · Proceedings - ACM IEEE Design Automation Conference · doi:10.1109/dac63849.2025.11133406 · Cited by 8
“…layer of 2 to 6 the loss weights are 0.8, 0.6, 0.4, 0.2, and 0.1 respectively, as depicted in Fig.1-b. 2) Curriculum Learning: As demonstrated in [21], employing a meaningful training sequence—beginning with simpler samples and progressing to more complex ones—can enhance model performance…”
InfluentialHwang, M., Rachim, V. P., Yoo, J., Lee, Y., & Park, S. (2025) · npj Digital Medicine · doi:10.1038/s41746-025-01994-4 · Cited by 13
SHISS: Supervised hashing with informative set selection
Ma, C., Gong, C., Gu, Y., Yang, J., & Feng, D. (2017) · Pattern Recognition Letters · doi:10.1016/j.patrec.2017.08.012 · Cited by 8
Deep Learning Approaches to Text Production
Narayan, S., & Gardent, C. (2020) · Synthesis lectures on human language technologies · doi:10.2200/s00979ed1v01y201912hlt044 · Cited by 22
Integrating Multiple Policies for Person-Following Robot Training Using Deep Reinforcement Learning
Dewa, C. K., & Miura, J. (2021) · IEEE Access · doi:10.1109/access.2021.3082136 · Cited by 11
“Nevertheless, instead of using fixed weights, our proposed weight-scheduled action smoothing follows the curriculum learning strategy [17] to schedule the weights during the training process so that the RL agent (the robot) can explore the environment sufficiently for finding the right and the left…”
MethodInfluentialAn analytical theory of curriculum learning in teacher–student networks*
Saglietti, L., Mannelli, S. S., & Saxe, A. (2022) · Journal of Statistical Mechanics Theory and Experiment · doi:10.1088/1742-5468/ac9b3c · Cited by 26
“In the following, we revisit a prototypical model of curriculum learning from [12] that finds correspondence to the fading literature [6] as highlighted in Sec.”
BackgroundMethodInfluentialGeneralization in Multimodal Language Learning from Simulation
Eisermann, A., Lee, J. H., Weber, C., & Wermter, S. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9534275 · Cited by 10
“We used curriculum learning to encourage the models to learn words more independently, which also led them to learn faster [20].”
MethodSituated Dialogue Learning through Procedural Environment Generation
Ammanabrolu, P., Jia, R., & Riedl, M. O. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-long.557 · Cited by 14
“Curriculum Learning Curriculums in reinforcement learning have traditionally been used to set goals of steadily increasing difficulty for an agent (Bengio et al., 2009; Schmidhuber, 2013).”
Hierarchical Curriculum Learning for AMR Parsing
Wang, P., Chen, L., Liu, T., Dai, D., Cao, Y., Chang, B., & Sui, Z. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-short.37 · Cited by 9
“Humans usually adapt to difficult tasks by dealing with examples gradually from easy to hard, i.e., Curriculum Learning (Bengio et al., 2009; Platan-ios et al., 2019; Su et al., 2021; Xu et al., 2020a (Cai and Lam, 2019).”
BackgroundHAR: Hardness Aware Reweighting for Imbalanced Datasets
Duggal, R., Freitas, S., Dhamnani, S., Chau, D. H., & Sun, J. (2021) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata52589.2021.9671807 · Cited by 17
“The concept of example hardness is central to a variety of successful learning techniques including those from curriculum learning [32], [33] (which deals with ordering the training set in increasing order of hardness), hard negative mining [13], [22], [34] (which deals with identifying false positive samples in order…”
BackgroundRobust Reinforcement Learning via Genetic Curriculum
Song, Y., & Schneider, J. (2022) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra46639.2022.9812420 · Cited by 20
“Organizing training data to gradually introduce more complex concepts, [20], curricular learning has shown success in supervised learning tasks [21]– [25] as well as various RL tasks [26]–[33].”
Adaptive Face Recognition Using Adversarial Information Network
Wang, M., & Deng, W. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3189830 · Cited by 9
“We additionally apply the easy-to-hard strategy of CL [48] in our AIN.”
BackgroundMethodClean or Annotate: How to Spend a Limited Data Collection Budget
Chen, D., & Bowman, Z. Y. A. S. R. (2022) · Workshop on Deep Learning Approaches for Low-Resource NLP (DeepLo) · doi:10.18653/v1/2022.deeplo-1.17 · Cited by 21
“Curriculum learning also selects examples for training based on model uncertainty (Bengio et al., 2009) and diversity maximization (Jiang et al., 2014).”
BackgroundNeuDep: neural binary memory dependence analysis
Pei, K., She, D., Wang, M., Geng, S., Xuan, Z., David, Y., Yang, J., Jana, S., & Ray, B. (2022) · ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE) · doi:10.1145/3540250.3549147 · Cited by 14
“For 𝑇 𝑖 whose aligned 𝐼 𝑖 is not executed, we assign the value 𝑏 𝑗 = 0x100 , ∀ 𝑗 ∈ [ 1 , 8 ] .”
BackgroundMethodInfluentialTowards Sustainable Compressive Population Health: A GAN-based Year-By-Year Imputation Method
Feng, Y., Wang, J., Wang, Y., & Chu, X. (2022) · ACM Transactions on Computing for Healthcare · doi:10.1145/3571159 · Cited by 8
“The authors in [4] propose a Curriculum Learning approach that advocates starting the model with easy samples and gradually progressing to complex samples and knowledge.”
BackgroundSha, X., & Diao, N. (2022) · Measurement · doi:10.1016/j.measurement.2022.112283 · Cited by 8
Population-coded Spiking Neural Network with Reinforcement Learning for Mapless Navigation
Xu, R., Wu, Y., Qin, X., & Zhao, P. (2022) · International Conference on Cyber-Physical Social Intelligence (ICCSI) · doi:10.1109/iccsi55536.2022.9970598 · Cited by 8
“Curriculum training has been demonstrated to allow better generalization and lead to faster convergence [18].”
BackgroundDiscriminative Mutual Learning for Multi-target Domain Adaptation
Wang, J., Zhong, C., Feng, C., Zhang, Y., Sun, J., & Yokota, Y. (2022) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr56361.2022.9956437 · Cited by 8
“Among the different sequences of transferring, easy-to-hard is recommended by curriculum learning [31], [18], [26].”
BackgroundTemporal action detection with dynamic weights based on curriculum learning
Chen, Y., Jiang, H., Xiao, J., Ding, L., & Gu, Q. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.12.049 · Cited by 8
Attention Augmented Face Morph Detection
Aghdaie, P., Soleymani, S., Nasrabadi, N. M., & Dawson, J. M. (2023) · IEEE Access · doi:10.1109/access.2023.3254539 · Cited by 11
“We follow the rules set forth by the curriculum learning paradigm [16], [37], [38] to incrementally incorporate channel-wise and spatial feature selection/refinement modules.”
MethodCLIP: Train Faster with Less Data
Khan, M. A., Hamila, R., & Menouar, H. (2023) · IEEE International Conference on Big Data and Smart Computing (BigComp) · doi:10.1109/bigcomp57234.2023.00014 · Cited by 12
“The idea of CL was first introduced long ago in [5] and formalized in [6].”
BackgroundA Signed Subgraph Encoding Approach via Linear Optimization for Link Sign Prediction
Fang, Z., Tan, S., & Wang, Y. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3280924 · Cited by 14
“We adopt the rectified linear unit (ReLU) as the activation function for all hidden layers and the Adam update rule for optimization [46] with mini-batch size 512 and learning rate 0.001.”
MethodInfluentialWei, Y., Zhang, H., Wang, Y., & Huang, C. (2023) · IEEE Access · doi:10.1109/access.2023.3297095 · Cited by 10
“learning and named them as curriculum learning [14].”
BackgroundReliaMatch: Semi-Supervised Classification with Reliable Match
Jiang, T., Chen, L., Chen, W., Meng, W., & Qi, P. (2023) · Applied Sciences · doi:10.3390/app13158856 · Cited by 10
Single/Multi-Source Black-Box Domain Adaption for Sensor Time Series Data
Ren, L., & Cheng, X. (2023) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2023.3300832 · Cited by 9
“We apply the curriculum learning strategy [51] to make the value of dynamic threshold T t gradually decrease with the process of model training with the step t .”
MethodArchitecture, Dataset and Model-Scale Agnostic Data-free Meta-Learning
Hu, Z., Shen, L., Wang, Z., Liu, T., Yuan, C., & Tao, D. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00747 · Cited by 17
“Curriculum learning (CL) [2, 35, 39] aims to train a model in a meaningful order during loading training data, from easy to hard.”
Dynamic Curriculum Learning via In-Domain Uncertainty for Medical Image Classification
Li, C., Li, M., Peng, C., & Lovell, B. C. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43904-9_72 · Cited by 10
Multi-label contrastive hashing
Wei, Z., Jin, K., Zhang, Z., & Zhou, X. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110239 · Cited by 11
HOPE: A Hierarchical Perspective for Semi-Supervised 2D-3D Cross-Modal Retrieval
Zhang, F., Zhou, H., Hua, X., Chen, C., & Luo, X. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3412760 · Cited by 16
“Inaddition,weretrievepre-cise and well-balanced unlabeled samples using anchors with class-specific thresholds adjusted by curriculum learning [20].”
MethodAdaCLF: An Adaptive Curriculum Learning Framework for Emotional Support Conversation
Tu, G., Niu, T., Xu, R., Liang, B., & Cambria, E. (2024) · IEEE Intelligent Systems · doi:10.1109/mis.2024.3411369 · Cited by 9
Dynamic Color Assignment for Hierarchical Data
Chen, J., Yang, W., Jia, Z., Xiao, L., & Liu, S. (2024) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2024.3456386 · Cited by 13
“H C ] 20 J u l 2024 process towards more promising regions in the solution space [2].”
BackgroundMethodLarge Language Model-Driven Curriculum Design for Mobile Networks
Erak, O., Alhussein, O., Naser, S. A., Alabbasi, N., Mi, D., & Muhaidat, S. (2024) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc62479.2024.10682015 · Cited by 8
“Bengio et al. first introduced this concept, illustrating how it can mimic human learning processes to benefit machine learning models [9].”
BackgroundKovač, G., Portelas, R., Dominey, P. F., & Oudeyer, P. (2024) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2024.1396359 · Cited by 46
Bieger, J., Þórisson, K. R., & Garrett, D. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-09274-4_1 · Cited by 12
“This technique has recently been utilized in reinforcement learning [12][20], unsupervised dependency parsing [21] and deep learning [4][3][10][13].”
BackgroundMethodPredictive feature selection for genetic policy search
Loscalzo, S., Wright, R., & Yu, L. (2014) · Autonomous Agents and Multi-Agent Systems · doi:10.1007/s10458-014-9268-y · Cited by 8
“An open future direction is to determine a best progression of feature subsets, and may be able to leverage concepts and techniques from the emerging field of Curriculum Learning [3].”
BackgroundTeaching Deep Learners to Generalize
Aggarwal, C. C. (2018) · Springer eBooks · doi:10.1007/978-3-319-94463-0_4 · Cited by 10
Learning Physically Based Humanoid Climbing Movements
Naderi, K., Babadi, A., & Hämäläinen, P. (2018) · Computer Graphics Forum · doi:10.1111/cgf.13513 · Cited by 8
“Thus, we adopted the curriculum learning [BLCW09] approach of progressing from simple to complex moves to minimize overall optimization failure rate.”
MethodAn Evolution Strategy with Progressive Episode Lengths for Playing Games
Fuks, L., Awad, N. H., Hutter, F., & Lindauer, M. (2019) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2019/172 · Cited by 13
“Furthermore, our work is closely related to curriculum learning [Bengio et al., 2009; Jiang et al., 2015], since we also organize experiences in a meaningful order which gradually introduces more concepts and more complex concepts.”
MethodCurriculum-based Teacher Ensemble for Robust Neural Network Distillation
Panagiotatos, G., Passalis, N., Iosifidis, A., Gabbouj, M., & Tefas, A. (2019) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco.2019.8903112 · Cited by 9
“This behavior can be also explained if we draw an analogy with the way that humans learn through curriculum-based learning [13].”
ResultSimilarity-Aware Network Embedding with Self-Paced Learning
Huang, C., Shi, B., Zhang, X., et al. (2019) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3357384.3358163 · Cited by 8
Stochastic batch size for adaptive regularization in deep network optimization
Nakamura, K., Soatto, S., & Hong, B. (2022) · Pattern Recognition · doi:10.1016/j.patcog.2022.108776 · Cited by 10
“Importance sampling aims to limit the estimation of gradients to informative samples for efficiency and variance reduction by considering manual selection [8], temporal history of gradient [56, 72], maximizing the diversity of losses [27,87], or largest changes in parameters [10, 20, 41, 42, 47, 81].”
BackgroundImproving Sequence-to-Sequence Pre-training via Sequence Span Rewriting
Zhou, W., Ge, T., Xu, C., Xu, K., & Wei, F. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.45 · Cited by 16
“the SSR training examples with curriculum learning (Bengio et al., 2009) according to their difficulties.”
BackgroundMethodScheduling the NASA Deep Space Network with Deep Reinforcement Learning
Goh, E. Y., Venkataram, H. S., Hoffmann, M., Johnston, M. D., & Wilson, B. D. (2021) · IEEE Aerospace Conference (AERO) · doi:10.1109/aero50100.2021.9438519 · Cited by 10
“To that end, we plan to integrate curriculum learning [30] and scale the training examples gradually using curriculum-based training strategies.”
MethodLi, X., Li, Q., Wang, W., et al. (2021) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2021.3073936 · Cited by 8
CAPABILITY ITERATION NETWORK FOR ROBOT PATH PLANNING
Nie, B., Gao, Y., Mei, Y., & Gao, F. (2021) · International Journal of Robotics and Automation · doi:10.2316/j.2021.206-0598 · Cited by 9
“To speed up convergence of the capability module, curriculum learning [39] is needed.”
BackgroundCan Transformer Language Models Predict Psychometric Properties?
Laverghetta, A., Nighojkar, A., Mirzakhalov, J., & Licato, J. (2021) · Joint Conference on Lexical and Computational Semantics · doi:10.18653/v1/2021.starsem-1.2 · Cited by 15
“This allowed them to create a dynamic curriculum learning (Bengio et al., 2009) algorithm, which achieved superior performance to the same models trained”
MethodProgressive Stage-wise Learning for Unsupervised Feature Representation Enhancement
Li, Z., Liu, C., Yuille, A., Ni, B., Zhang, W., & Gao, W. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00964 · Cited by 9
“As curriculum learning [5] suggests, when dealing with a complex learning target, learning things progressively can be very useful.”
BackgroundCurriculum Learning based approaches for robust end-to-end far-field speech recognition
Ranjan, S., & Hansen, J. H. L. (2021) · Speech Communication · doi:10.1016/j.specom.2021.06.003 · Cited by 10
CurrMG: A Curriculum Learning Approach for Graph Based Molecular Property Prediction
Gu, Y., Zheng, S., & Li, J. (2021) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm52615.2021.9669478 · Cited by 14
NPT-Loss: Demystifying Face Recognition Losses With Nearest Proxies Triplet
Khalid, S. S., Awais, M., Feng, Z.-H., Chan, C.-H., Farooq, A., Akbari, A., & Kittler, J. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3162705 · Cited by 24
Supervised Attention Network for Arbitrary-Shaped Text Detection in Edge-Fainted Noisy Scene Images
Soni, A., Dutta, T., Nigam, N., Verma, D., & Gupta, H. P. (2022) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2022.3153557 · Cited by 8
“We utilize the curriculum learning technique [50] for training of our SESANet slowly starting with simple to complex samples meticulously.”
MethodDrop Loss for Person Attribute Recognition With Imbalanced Noisy-Labeled Samples
Yan, Y., Xu, Y., Xue, J., Lu, Y., Wang, H., & Zhu, W. (2022) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2022.3173356 · Cited by 9
“Note that the above learning procedure is different from curriculum learning [50], where the training process pays attention to easy samples at the beginning and gradually focuses on difficult samples in subsequent training.”
BackgroundMuscle‐driven virtual human motion generation approach based on deep reinforcement learning
Qin, W., Tao, R., Sun, L., & Dong, K. (2022) · Computer Animation and Virtual Worlds · doi:10.1002/cav.2092 · Cited by 11
“τ n = −kp ( qn+1 − qn+1 tar ) − kdq̇, (13)”
BackgroundInfluentialOptimizing Constrained Guidance Policy With Minimum Overload Regularization
Luo, W., Chen, L., Liu, K., Gu, H., & Lü, J. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3163463 · Cited by 8
A single neural cellular automaton for body-brain co-evolution
Pontes-Filho, S., Walker, K., Najarro, E., Nichele, S., & Risi, S. (2022) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3520304.3529002 · Cited by 8
Curriculum Learning-Based Approaches for End-to-End Gas Recognition
Zhang, C., Wang, W., Pan, Y., & Zhai, S. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3205901 · Cited by 9
“The cur-44 riculum is an important feature of human learning, which usually start with the simple concepts in the learning process, then gradually generalizing to more complex concepts; this has been applied to DNN models and is referred to as curriculum learning (CL) [10].”
MethodAssessing evidence accumulation and rule learning in humans with an online game
Do, Q. H., Li, Y., Kane, G. A., McGuire, J. T., & Scott, B. B. (2022) · Journal of Neurophysiology · doi:10.1152/jn.00124.2022 · Cited by 7
Accelerating Self-Supervised Learning via Efficient Training Strategies
Kocyigit, M. T., Hospedales, T. M., & Bilen, H. (2023) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00561 · Cited by 17
“Curriculum learning [2] has been used to accelerate convergence for deep learning methods and when applying it to self-supervised learning, augmentation strength and image resolution are the most relevant parameters.”
MethodDouble Deep Reinforcement Learning
Kiefer, J., & Dorer, K. (2023) · IEEE International Conference on Autonomous Robot Systems and Competitions · doi:10.1109/icarsc58346.2023.10129640 · Cited by 8
UnShadowNet: Illumination Critic Guided Contrastive Learning for Shadow Removal
Dasgupta, S., Das, A., Yogamani, S. K., Das, S., Eising, C., Bursuc, A., & Bhattacharya, U. (2023) · IEEE Access · doi:10.1109/access.2023.3305576 · Cited by 7
“[69] P. Bachman, R. Devon Hjelm, and W. Buchwalter, ‘‘Learning representations by maximizing mutual information across views,’’ 2019, arXiv:1906.00910.”
BackgroundInfluentialProgressive Transformation Learning for Leveraging Virtual Images in Training
Shen, Y., Lee, H., Kwon, H., & Bhattacharyya, S. S. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00087 · Cited by 15
“Curriculum learning [6, 15, 26, 45], which continuously raises the level of training from easy to difficult samples, also falls into this category.”
Bai, J., Guo, H., Liu, J., Yang, J., Liang, X., Yan, Z., & Li, Z. (2023) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3583780.3614901 · Cited by 19
“Inspired by curriculum learning [2], we propose a curriculum knowledge distillation mechanism for better distilling the knowledge from GPE to ranker.”
BackgroundMethodInfluentialArtificial Intelligence and Machine Learning in Drug Discovery and Development
Garg, S., Arora, K., Singh, S., & Nagarajan, K. (2023) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3693-0368-9.ch003 · Cited by 13
When Do Curricula Work in Federated Learning?
Vahidian, S., Kadaveru, S., Baek, W., Wang, W., Kungurtsev, V., Chen, C., Shah, M., & Lin, B. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00469 · Cited by 15
Müller, A., Grumbach, F., & Kattenstroth, F. (2024) · IEEE Access · doi:10.1109/access.2024.3355269 · Cited by 13
“To guide the agent in the learning process, we use curriculum learning [32], [33], i.e. the agent is exposed to a sequence of tasks with increasing complexity.”
MethodReal-Time 4K Super-Resolution of Compressed AVIF Images. AIS 2024 Challenge Survey
Conde, M. V., Lei, Z., Li, W., Katsavounidis, I. X., Timofte, R., Yan, M., Liu, X., Wang, Q., Ye, X., Du, Z., et al. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00592 · Cited by 31
“We also applied curriculum learning [2] to efficiently learn lightweight models.”
Zhang, J., Zhang, D., Ma, W., & Jiao, L. (2018) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2018.2854847 · Cited by 8
“[10], which suggests learning easy samples first, and then, complex samples are added to the learning algorithm gradually.”
BackgroundFrustratingly Easy Trade-off Optimization Between Single-Stage and Two-Stage Deep Object Detectors
Soviany, P., & Ionescu, R. T. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-11018-5_33 · Cited by 12
“[1] have explored easy-to-hard strategies to train machine learning models, showing that machines can also benefit from learning by gradually adding more difficult examples.”
BackgroundMethodArtifact-driven sampling schemes for robust female pelvis CBCT segmentation using deep learning
Gerken, A., Dicken, V., Klein, J., Morgaś, T., Haas, B., & Hahn, H. K. (2019) · SPIE Medical Imaging: Computer-Aided Diagnosis · doi:10.1117/12.2512727 · Cited by 8
Detecting Spam Images with Embedded Arabic Text in Twitter
Imam, N. H., & Vassilakis, V. G. (2019) · International Conference on Document Analysis and Recognition Workshops (ICDARW) · doi:10.1109/icdarw.2019.50107 · Cited by 10
“In this approach, which was first introduced by [44], the training process starts with a”
BackgroundMeta Transfer Learning for Adaptive Vehicle Tracking in UAV Videos
Song, W., Li, S., Guo, Y., Li, S., Zhao, Q., Qin, H., & Zhao, Q. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-37731-1_62 · Cited by 7
Simple and automated negative sampling for knowledge graph embedding
Zhang, Y., Yao, Q., & Chen, L. (2021) · The VLDB Journal · doi:10.1007/s00778-020-00640-7 · Cited by 18
“The main idea of self-paced learning (or curriculum learning) [3,31] is to pick up easy samples first and then gradually switch to harder ones.”
BackgroundWu, Y., Chen, L., & Merhof, D. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-65414-6_16 · Cited by 11
“Our idea was indirectly inspired by the easy task first concept behind curriculum learning [1].”
InfluentialProgressive Co-Teaching for Ambiguous Speech Emotion Recognition
Yin, Y., Gu, Y., Yao, L., Zhou, Y., Liang, X., & Zhang, H. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414494 · Cited by 10
Improving the Accuracy of Early Exits in Multi-Exit Architectures via Curriculum Learning
Bakhtiarnia, A., Zhang, Q., & Iosifidis, A. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9533875 · Cited by 14
“Curriculum learning [12] is a training strategy for neural networks that has been shown to improve the final accuracy of a network in certain cases.”
BackgroundMethod
Metadata
- Authors
- Bengio, Y., Louradour, J., Collobert, R., & Weston, J.
- Affiliations
- Université de Montréal (CA)
- NEC Laboratories America, Princeton, NJ
- ORCID
- Bengio, Yoshua
- Venue
- Proceedings of the 26th Annual International Conference on Machine Learning
- Publisher
- ACM
- Published
- 2009-06-14
- Pages
- 41-48
- OpenAlex
- W2296073425
- Semantic Scholar
- 8de174ab5419b9d3127695405efd079808e956e8
- Cited by
- 5,366
- References
- 40
- Citations by year
- 2026: 526 · 2025: 636 · 2024: 551 · 2023: 591 · 2022: 536 · 2021: 711 · 2020: 565 · 2019: 442 · 2018: 293 · 2017: 198 · 2016: 103 · 2015: 63 · 2014: 41 · 2013: 40 · 2012: 33
- FWCI
- 16.17
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- International Conference on Machine Learning (publishedVersion)
- Queen Mary Research Online (Queen Mary University of London) (submittedVersion)
- Funders
- Canadian Institute for Advanced Research
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Stochastic Gradient Optimization Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Keywords
- curriculum learning, training order, non-convex optimization, deep neural network, stochastic neural networks, deterministic neural networks, generalization, convergence rate, numerical continuation, global optimization
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.
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.