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 3,301–3,400 of 5,240
Preview-Based Category Contrastive Learning for Knowledge Distillation
Ding, M., Wu, J., Dong, X., Li, X., Qin, P., Gan, T., & Nie, L. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3541067 · Cited by 5
“Towards this issue, the curriculum learning methods [51], [64] propose to pre-calculate the sample difficulty and then make the student learn from the easy samples as well as filter out these hard samples.”
BackgroundMethodProgressive Subband Modeling for Artifacts-free Speech Super-resolution
Kim, D., & Chang, J. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10889911 · Cited by 2
“Progressive curriculum learning (PCL) [30], on the other hand, refers to the task in which the complexity of tasks presented to the model gradually increases as training progresses.”
Flex: Fast, Accurate DNN Inference on Low-Cost Edges Using Heterogeneous Accelerator Execution
Sen, T., Shen, H., & Iyer, A. P. (2025) · European Conference on Computer Systems · doi:10.1145/3689031.3696067 · Cited by 5
“To reduce the overhead RL training and enhance model generalization across diverse architectures, we use curriculum learning [30].”
Lin, L., Lin, C., Nie, L., et al. (2025) · The Visual Computer · doi:10.1007/s00371-025-03889-7 · Cited by 2
Grey-Box Adversarial Attack on Communication in Communicative Multi-Agent Reinforcement Learning
Ma, X., & Li, W.-J. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3560203 · Cited by 2
“Following [20], we also utilize curriculum learning [55] to make the training process easier.”
An All-in-One Approach for Accelerated Cardiac MRI Reconstruction
Hamedani, K. A., Razizadeh, N., Nabavi, S., & Moghaddam, M. E. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-87756-8_45 · Cited by 2
“Curriculum learning [2] starts with lower acceleration factors, allowing the model to grasp the fundamentals of reconstruction in an easy-to-hard manner.”
Fang, B., Liu, Z., & Wu, Q. (2025) · International Symposium on System Security, Safety, and Reliability (ISSSR) · doi:10.1109/isssr65654.2025.00015 · Cited by 2
“Given that physical constraints have the potential to disrupt the convergence of the model during the initial training phases, a Curriculum Learning [24] strategy is implemented to dynamically modulate the weight of the physical regularization term.”
Yang, Z., Tullo, D., & Rabbany, R. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3711896.3737271 · Cited by 8
Dong, J., Wei, J., Hu, X., Dong, Z., Chen, F., Hu, X. Y., & Qi, J. (2025) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3574417 · Cited by 5
“Initially introduced by Bengio et al. [17], CL implements training from easy to hard data.”
Hsieh, I.-T., & Wu, C. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3597438 · Cited by 3
“ThecomparisonincludesCL[42],MAML[37],DA[36],[30], [57], [31], and similar model architecture [58], [53], with results presented in Table X.”
CD-GANs : Conditional dynamic training with GANs for enhanced post-gadolinium glioblastoma MRIs
Ahriz, A., Saouli, R., Raad, L., & Cousty, J. (2025) · Magnetic Resonance Imaging · doi:10.1016/j.mri.2025.110509 · Cited by 2
UniROS: ROS-Based Reinforcement Learning Across Simulated and Real-World Robotics
Kapukotuwa, J., Lee, B., Devine, D. M., & Qiao, Y. (2025) · Sensors · doi:10.3390/s25185679 · Cited by 5
“Furthermore, incorporating techniques such as Curriculum Learning [32], where the agent maneuvers through complex tasks by breaking them down into simpler subtasks and gradually increasing the task difficulty during learning, can help train robots directly without a simulation environment [33].”
Transformer-Based World Interaction Modeling for Humanoid Locomotion Control
Zheng, H., Cheng, Y., Liu, H., Li, J., Li, Y., Ye, L., & Liu, H. (2025) · IEEE International Conference on Automation Science and Engineering (CASE) · doi:10.1109/case58245.2025.11164052 · Cited by 3
“The loss function primarily consists of a policy loss L clip and a value function loss L value The overall training loss is defined as We adopt curriculum learning [44] for training.”
Tao, J., Shen, T., Zhao, D., & Zhang, F. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3662620 · Cited by 2
“Curriculum learning [27] is employed to enhance training stability and mitigate the risk of converging to suboptimal lo-cal minima.”
Miller, B., & Ziemiański, L. (2025) · Applied Sciences · doi:10.3390/app151910783 · Cited by 2
Enhancing peptide identification in metaproteomics through curriculum learning in deep learning
Feng, S., Zhang, B., Wang, H., Xiong, Y., Tian, A., Yuan, X., Pan, C., & Guo, X. (2025) · Nature Communications · doi:10.1038/s41467-025-63977-z · Cited by 2
“The training process employs a curriculum learning strategy 29 to enhance model performance and accelerate convergence.”
Wen, Z., Tan, H., Zhao, Y., Zhang, H., Li, P., Chen, X., & Gao, B. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3611296 · Cited by 2
“Therefore, we introduce curriculum learning [43] in the training of MO-HRL.”
Power Efficient Joint Channel Coding and Frequency Modulation With Deep Learning
Chang, B., Shen, Y., Wang, C., & Kim, H.-S. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3620387 · Cited by 2
“Inspired by curriculum learning [35] and domain adaptation techniques [36], we empirically find that beginning with simpler data distributions significantly improves training stability and convergence.”
Enhancing Diffusion-based Dataset Distillation via Adversary-Guided Curriculum Sampling
Zou, L., Chen, G., Chen, Y., & Zhang, M. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209069 · Cited by 3
Zhu, X., Qi, J., Hua, J., Li, T., Yang, L., Dong, Z., & Sun, Y. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3631089 · Cited by 3
“Dynamic factor β : In order to mitigate the potential negative e ff ects caused by focal loss, the idea of curriculum learning [38] is introduced.”
Karzanov, D., Garzón, R., Terekhov, M. V., Gülçehre, Ç., Raffinot, T., & Detyniecki, M. (2025) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3768292.3770340 · Cited by 8
“Inspired by the principles of Curriculum Learning [3, 22], where the complexity of tasks gradually increases from simple to more real-world scenarios, we introduce a transaction cost scheduler that incrementally raises the training transaction cost at each step according to The costs are increased…”
Possible Principles for Aligned Structure Learning Agents
Costa, L. D., Gavenčiak, T., Hyland, D., Samiei, M., Dragos-Manta, C., Pattisapu, C., Razi, A., & Friston, K. J. (2025) · Neural Computation · doi:10.1162/neco.a.39 · Cited by 8
Liu, S., Huang, M., Yang, G., & Zhou, W. (2025) · Scientific Reports · doi:10.1038/s41598-025-24135-z · Cited by 2
‘It keeps me awake at night’: machine-learning pioneer on AI’s threat to humanity
Castelvecchi, D., & Thompson, B. S. (2025) · Nature · doi:10.1038/d41586-025-03686-1 · Cited by 2
Yang, G., Chen, Z., Guo, P., Ma, J., Huang, J., Jin, C., Yang, X., Zhao, K., Wang, Y., Gao, Q., et al. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1649452 · Cited by 2
Reward Training Wheels: Adaptive Auxiliary Rewards for Robotics Reinforcement Learning
Wang, L., Xu, T., Lu, Y., & Xiao, X. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247039 · Cited by 9
“Curriculum learning, first formalized by Bengio et al. [13], has become increasingly important in RL as a strategy to improve sample efficiency and final performance.”
Real-Time Reinforcement Learning for Dynamic Tasks with a Parallel Soft Robot
Avtges, J., Ketchum, J., Schlafly, M., Young, H., Kim, T., Pinosky, A., Truby, R. L., & Murphey, T. D. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246440 · Cited by 5
“To overcome the challenge of unstable learning performance in arbitrary-point balancing, we employ curriculum learning (CL) [42], a procedure in which a sequence of tasks enables the transfer of a learned representation from easier problems to more difficult ones.”
Safe reinforcement learning for vision-based robotic manipulation in human-centered environments
Khan, F., Feng, W., Wang, Z., Huang, T., Liu, X., Cui, Y., & Wang, W. (2025) · International Journal of Intelligent Robotics and Applications · doi:10.1007/s41315-025-00507-6 · Cited by 3
Progressive curriculum learning with Scale-Enhanced U-Net for continuous airway segmentation
Yang, B., Tian, Q., Liao, H., et al. (2025) · BMC Medical Imaging · doi:10.1186/s12880-025-02066-5 · Cited by 2
“To address intra-class imbalance and reduce interference in feature learning, we propose a progressive curriculum learning pipeline for airway segmentation, inspired by curriculum learning [14].”
InfluentialRen, R., Zhao, H., Wang, Y., Lyu, S., Wang, G., Cheng, G., Zhao, Q., & Ren, J. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3643913 · Cited by 2
“Furthermore, some researches [44], [45] propose that the process from easier samples to harder samples is beneficial to denoising the noisy data for harder samples.”
Mantovani, L. Q., & Schaub, H. (2026) · Journal of Aerospace Information Systems · doi:10.2514/1.i011697 · Cited by 5
González, A., Félix, R. A., Mena, L. J., Félix, V. G., Martínez-Peláez, R., Ostos, R., Velarde-Alvarado, P., & Ochoa-Brust, A. M. (2026) · Technologies · doi:10.3390/technologies14010044 · Cited by 5
“Curriculum learning progressively increases task complexity during training, enabling more efficient learning and improved generalization [44,45].”
Shao, S., Ding, X., Zhao, B., & Ye, P. (2026) · Scientific Reports · doi:10.1038/s41598-026-38622-4 · Cited by 7
“We adopt a curriculum learning approach that progressively increases prediction horizon length during training, beginning with single-step-ahead forecasts before advancing to longer horizons 49 .”
LLM-Driven Expressive Robot Motion Via Proxy-Based Optimization
Roy, L., Croft, E. A., Ramirez, A., & Kulić, D. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3666351 · Cited by 3
“These results align with established curriculum theories in HRI and education [28], which emphasize that the gradual introduction of complexity enhances learning.”
CurST-Net: Curriculum Learning Guided Spatial–Temporal Network for Traffic Flow Prediction
Chen, S., Chen, X., Huang, Y., Su, Y. C. F., Liu, G., & Fu, Y.-G. (2026) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2026.3661914 · Cited by 2
“CL is a training strategy that was first proposed by [38].”
Hu, R., Wang, Q., Yan, H., et al. (2026) · Energy · doi:10.1016/j.energy.2026.140539 · Cited by 2
QoE-Aware UAV Trajectory Planning via Reinforcement Learning for Post-Disaster Network
Phetxomphou, Z., Le, H. D., & Pham, A. T. (2025) · doi:10.23919/icetccom67395.2025.11416813 · Cited by 2
Li, H., Chu, H. K., & Sun, Y. (2026) · IEEE Transactions on Robotics · doi:10.1109/tro.2026.3677009 · Cited by 2
“This method is similar to curriculum learning strategies [64], where the model masters easier concepts (edge regions) to build a strong feature foundation before fully tackling the harder ones (central regions) and, thus, improve the overall robustness and accuracy for segmentation performance.”
Lv, J., Wu, Q., Liu, J., Yang, B., Li, Y., Xu, J., Meng, Y., Wei, L., Zhang, Z., Zou, Q., et al. (2026) · Advanced Science · doi:10.1002/advs.202523984 · Cited by 3
“We adopted a mild two-stage curriculum [ 41 ].”
CheMLT-F: multitask learning in biochemistry through transformer fusion
Mun, V., & Fazli, S. (2026) · Journal of Cheminformatics · doi:10.1186/s13321-026-01199-1 · Cited by 2
“For optimization, established multitask balancing schemes–uncertainty weighting, GradNorm, and multi-objective formulations–could improve stability, optionally paired with simple curricula or temper-ature-based sampling [71].”
Poudineh, M., Esmaeilzehi, A., & Ahmad, M. O. (2026) · IEEE Transactions on Image Processing · doi:10.1109/tip.2026.3685457 · Cited by 2
“To further clarify the above points, it has been shown in [23] that by using hard training samples in the training process of deep neural networks, their learning capabilities would be deteriorated, and hence, inferior performances are yielded.”
Towards Long-Horizon Vision-Language-Action System: Reasoning, Acting and Memory
Li, D., Zhang, Y., Cao, M., Donglai, L., Xie, W., Hui, T., Lin, L., Xie, Z., & Li, Y. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00643 · Cited by 14
“To effectively guide training, we employ a curriculum learning approach [3], where data is arranged in order from easy to difficult.”
NIDS-Mamba: Lightweight Network Intrusion Detection for IoT Sensor Networks via State Space Models
Ding, Z., Zheng, J., & Wu, X. (2026) · Sensors · doi:10.3390/s26092766 · Cited by 6
“Additionally, generative models such as GANs or diffusion models may be employed to synthesize high-fidelity minority attack flows [41], thereby improving the model’s ability to extract representations from small-sample anomalies.”
Xiao, L., Li, J., Chen, Y., Qiu, C., Wang, Q., Liao, X., & Chen, J. (2026) · Artificial Intelligence Chemistry · doi:10.1016/j.aichem.2026.100119 · Cited by 2
Kapwata, T., Webster, C., Chen, W. G. G., Suder, M. I., Jarvis, D., Vogt, A., Mahlangeni, N. T., Laban, T. L., Naidoo, N., Mudu, P., et al. (2026) · Environmental Monitoring and Assessment · doi:10.1007/s10661-026-15376-0 · Cited by 5
“The various stages we employed formed a “cur-riculum” (Bengio et al., 2009) which exposed the model to various tasks in sequence of increasing difficulty.”
Cells on Autopilot: Adaptive Cell (Re)Selection via Reinforcement Learning
Illian, M., Khalili, R., Rocha, A. A. D. A., et al. (2026) · International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt) · doi:10.23919/wiopt71098.2026.11568224 · Cited by 2
“To mitigate this, we employ a powerful technique called curriculum learning [17], starting with shorter episodes which provide a simpler environment.”
Towards Visual Training Set Generation Framework
Hůla, J., Perfilieva, I., & Muzaheed, A. A. M. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-59147-6_63 · Cited by 2
Leu, G., Tang, J., Lakshika, E., Merrick, K. E., & Barlow, M. (2017) · Asian Conference on Defence Technology · doi:10.1109/acdtj.2017.8259585 · Cited by 2
“In a more recent study [14], the authors refine the approach presented in [13], and use a similar curriculum learning method in conjunction with a Recurrent NN (RNN) with long short-term memory (LTSM) units which provides the memory substrate needed for potential application to skill transfer and generalisation to new…”
BackgroundMethodInvestigating the Compositional Structure of Deep Neural Networks
Craighero, F., Angaroni, F., Graudenzi, A., Stella, F. A., & Antoniotti, M. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-64583-0_30 · Cited by 6
“Again, to limit the scope of our investigation, we can distinguish four different sampling strategies: ( i ) curriculum learning [2], according to which it is preferable to start learning from easier to harder instances, also implemented in self-paced learning [9]; ( ii ) selecting only the hardest…”
BackgroundLocality Sensitive Batch Selection for Triplet Networks
Martin, K., Wiratunga, N., & Sani, S. (2020) · doi:10.1109/ijcnn48605.2020.9207538 · Cited by 2
“In response and inspired by research in curriculum learning [22], in particular self-paced learning [10], the proposed method in this paper uses latest network output to inform its clustering.”
MethodKodaka, I., & Saito, F. (2021) · IEEE International Workshop on Computational Intelligence and Applications (IWCIA) · doi:10.1109/iwcia52852.2021.9626020 · Cited by 2
Confidence-Guided Learning For Breast Density Classification
Park, S. Y., Sargent, D., & Richmond, D. (2021) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi48211.2021.9433965 · Cited by 2
“…four density classes on the test set using models trained with the standard batch generator and loss function (baseline 1, B1), curriculum learning [20] (baseline 2, B2) versus models trained with our proposed loss function (proposed 1, P1) and the loss function with the custom batch generator…”
MethodThe Seeing-Eye Robot Grand Challenge: Rethinking Automated Care
Mirsky, R., & Stone, P. (2021) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/sqki6856 · Cited by 15
Feedback Network With Curriculum Learning for Earthquake Event Classification
Min, J., Ku, B., & Ko, H. (2021) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2021.3097041 · Cited by 2
Chowdhury, S., Dasgupta, S., Das, S. K., & Bhattacharya, U. (2021) · Proceedings - International Conference on Image Processing · doi:10.1109/icip42928.2021.9506019 · Cited by 3
“Here, we have used curriculum learning [15] to enable the proposed AViS-Net architecture to detect multiple similar sound sources in the same scene.”
MethodRen, J., Zhang, Y., Zeng, Y., & Lan, Y. (2021) · IEEE International Conference on Real-time Computing and Robotics (RCAR) · doi:10.1109/rcar52367.2021.9517647 · Cited by 2
Single Cell Self-Paced Clustering with Transcriptome Sequencing Data
Zhao, P., Xu, Z., Chen, J., Ren, Y., & King, I. (2022) · International Journal of Molecular Sciences · doi:10.3390/ijms23073900 · Cited by 3
Gradual Domain Adaptation with Sample Transferability Exploitation for Person Re-Identification
Guo, Z.-K., Ma, B., Chang, H., & Chen, X. (2022) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme52920.2022.9859861 · Cited by 2
“The basic idea of curriculum learn ing is to "start small" [14], train the model on easier data first, and then gradually increase the difficulty of the training data.”
BackgroundMethodDelving into Light-Dark Semantic Segmentation for Indoor Scenes Understanding
Ying, X., Lang, B., Zheng, Z., & Chuah, M. C. (2022) · doi:10.1145/3552482.3556556 · Cited by 2
“This strategy has been shown to boost the speed and quality for optimization in non-convex cases [1].”
BackgroundHigh Myopia Detection Method On Fundus Images Based On Curriculum Learning
Zhao, H., Ma, Z., Guan, Y., et al. (2022) · IEEE Annual Computers, Software, and Applications Conference (COMPSAC) · doi:10.1109/compsac54236.2022.00227 · Cited by 2
Progressive Unsupervised Learning of Local Descriptors
Wang, W., Zhang, L., & Huang, H. (2022) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3503161.3547792 · Cited by 2
“Essentially, this also suggests that the RCA method inherently encapsulates a curriculum learning strategy [3] which encourages the CNN model to learn from training tuples in an easy-to-hard manner.”
MethodMulti-source inverse-curriculum-based training for low-resource dialogue generation
Cui, F., Di, H., Huang, H., Ren, H., Ouchi, K., Liu, Z., & Xu, J. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04190-z · Cited by 2
“Curriculum learning (CL) [23], imitating the human learning process, starts with easy data and then gradually learns the harder ones, which can help decrease training time and sometimes improve generalization.”
BackgroundTummala, B. M., & Barpanda, S. S. (2023) · Bulletin of Electrical Engineering and Informatics · doi:10.11591/eei.v12i3.4676 · Cited by 3
Bidirectional Sentence Ordering with Interactive Decoding
Bai, G., He, S., Liu, K., & Zhao, J. (2023) · ACM Transactions on Asian and Low-Resource Language Information Processing · doi:10.1145/3561510 · Cited by 2
“Similar to curriculum learning [5], easy-first decoding proposes to output sequences with an easy-to-hard direction as a human.”
BackgroundInference and Learning for Generative Capsule Models
Nazábal, A., Tsagkas, N., & Williams, C. K. I. (2023) · Neural Computation · doi:10.1162/neco_a_01564 · Cited by 4
“One advantage of the interpretable nature of the GCM is that one can use a “curriculum learning” approach (Bengio et al., 2009), where individual models can first be learned, and then composed together for inference in more complex situations. For a given object, a key issue is the learning of the decomposition into p…”
MethodInfluentialDeepADMR: A Deep Learning based Anomaly Detection for MANET Routing
Yahja, A., Kaviani, S., Ryu, B., Kim, J., & Larson, K. A. (2022) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom55135.2022.10017842 · Cited by 6
“Despite the generalization ability of learned DRL-based policy, however, applying these algorithms to different environments necessitates re-learning to attain similar performance, with methods such as curriculum learning [8], data augmentation [9], and policy distillation [10].”
MethodLearning Sentinel-2 Reflectance Dynamics for Data-Driven Assimilation and Forecasting
Frion, A., Drumetz, L., Tochon, G., Mura, M. D., & Bey, A. A. E. (2023) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco58844.2023.10289879 · Cited by 2
“Such a procedure is related to curriculum learning [12], which we believe to be crucial when learning difficult physics-related tasks (see [13] for a recent survey).”
BackgroundTask-Covariant Representations for Few-Shot Learning on Remote Sensing Images
Zhang, L., Tian, Z., Tang, Y., & Jiang, Z. (2023) · Mathematics · doi:10.3390/math11081930 · Cited by 2
EMC2-Net: Joint Equalization and Modulation Classification Based on Constellation Network
Ryu, H., & Choi, J. J. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10096687 · Cited by 3
“In order to train the classifier in a robust manner, inspired by [18], we apply noise-curriculum based on SNR values in which con-Authorized stellation points of the highest SNR are given first then those of low SNR are gradually added to the dataset as training goes by.”
MethodAI Algorithm for Image Enhancement
Rane, M., Kulkarni, M., Dalvi, A., et al. (2023) · International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies (ICAECT) · doi:10.1109/icaect57570.2023.10117819 · Cited by 2
“Using extensive PSNR techniques for redesigning of large up scaling factors [4] are, more photo-possible by a considerable margin than redesigning obtained to the fore confess-of the-art reference method [12].”
BackgroundMethodA Curriculum Learning Approach for Multi-Domain Text Classification Using Keyword Weight Ranking
Yuan, Z., Li, Y., Li, Y., et al. (2023) · Electronics · doi:10.3390/electronics12143040 · Cited by 2
Nie, L., Qiu, Y., Meng, F., et al. (2023) · Proceedings - IEEE International Parallel and Distributed Processing Symposium · doi:10.1109/ipdps54959.2023.00051 · Cited by 2
“Curriculum learning [17] mimics the education processes of humans in the real world, where learning problems come in a sequence from easy to difficult.”
BackgroundMethodIn-sample Curriculum Learning by Sequence Completion for Natural Language Generation
Jia, Q. F., Liu, Y., Tang, H., & Zhu, K. Q. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.666 · Cited by 8
Cai, W., Huang, L., & Zou, Z. (2023) · Lecture notes in civil engineering · doi:10.1007/978-3-031-34593-7_17 · Cited by 2
Fast Fine-Tuning Using Curriculum Domain Adaptation
Shen, L., Amara, I., Li, R., Meyer, B. H., Gross, W. J., & Clark, J. J. (2023) · Canadian Conference on Computer and Robot Vision · doi:10.1109/crv60082.2023.00045 · Cited by 4
“[3], aims to improve the performance of deep models by mimicking the human learning process.”
BackgroundResultInfluentialHighly Crowd Detection and Counting Based on Curriculum Learning
Fotia, L., Percannella, G., Saggese, A., & Vento, M. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44240-7_2 · Cited by 3
SWP:A Sliding Window Prompt for Emotion Recognition in Conversation
Zhao, H., Chen, Y., Xie, J., & Li, K. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44207-0_25 · Cited by 2
An Efficient Neural Network Learning Algorithm for Printed Spiral Coil (PSC) Impedance Prediction
Kim, J., & Song, E. (2023) · International Symposium on Electromagnetic Compatibility - EMC Europe (EMC Europe) · doi:10.1109/emceurope57790.2023.10274210 · Cited by 2
“To overcome this, a learning strategy that prioritizes training data with complex patterns is needed [15], [16].”
BackgroundTime-Series Prediction using Nature-Inspired Small Models and Curriculum Learning
Bothe, S., Farooq, H., Forgeat, J., & Čyras, K. (2023) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc56721.2023.10294056 · Cited by 4
Dy-KD: Dynamic Knowledge Distillation for Reduced Easy Examples
Lin, C., Jiang, N., Tang, J., Huang, X., & Wu, W. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-8148-9_18 · Cited by 3
Deep learning based active image steganalysis: a review
Bedi, P., Singhal, A., & Bhasin, V. (2023) · International Journal of Systems Assurance Engineering and Management · doi:10.1007/s13198-023-02203-9 · Cited by 5
“DL techniques like curriculum learning or transfer learning (Bengio et al. 2009) needs to be explored to tackle the problem of low embedding capacity in covert communication.”
HGLP: Hierarchical Solver for Combined Task Assignment and Path Finding
Gao, J., Ye, Z., Li, X., et al. (2023) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc58219.2023.10326804 · Cited by 2
“This paper adopts the curriculum learning [22].”
Empower Nested Boolean Logic via Self-Supervised Curriculum Learning
Wu, H., Liu, L., Zhao, H. Y., & Zhang, M. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.847 · Cited by 12
Car Full View Dataset: Fine-Grained Predictions of Car Orientation from Images
Cǎtrunǎ, A., Betiu, P., Tertes, E., Ghita, V., Rǎdoi, E., Mocanu, I. G., & Dascălu, M. (2023) · Electronics · doi:10.3390/electronics12244947 · Cited by 5
Enhancing Offline Coverage Path Planning with Deep Reinforcement Learning
Lian, Y., Yang, S., Jin, S., et al. (2023) · doi:10.1109/mlccim60412.2023.00056 · Cited by 2
de Oliveira, I. R. L., de Carvalho, K. B., & Brandao, A. S. (2023) · IEEE International Conference on Industry Applications · doi:10.1109/induscon58041.2023.10374605 · Cited by 2
“Curriculum learning proposes to train the agent in progressively harder scenarios, in this manner, the starting task can guide the learner to grasp the basic concepts of the situation and increase the success rate of the agent in the more complex situations [30].”
BackgroundPersistent-Transient Duality: A Multi-mechanism Approach for Modeling Human-Object Interaction
Tran, H., Le, V., Venkatesh, S., & Tran, T. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00904 · Cited by 2
“This way of training can be thought of as a curriculum learning technique [5] used in previous works [1, 25] where we initially train the model with an easy problem, then increase the difficulty of the problem in the later epochs.”
MethodA Multi-UAV Cooperative Tracking Strategy Based on Deep Reinforcement Learning
Jiang, H., Xu, R., Wu, Y., & Xue, L. (2023) · Youth Academic Annual Conference of Chinese Association of Automation (YAC) · doi:10.1109/yac59482.2023.10401834 · Cited by 2
MAPPO-Based Optimal Reciprocal Collision Avoidance for Autonomous Mobile Robots in Crowds
Liu, Z., Yao, C., Na, W., Liu, C., & Chen, Q. (2023) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc53992.2023.10394332 · Cited by 2
Crowd Density Estimation Based on Multi-Layer Feature Maps and Curriculum Learning
Xiao, J., Shu, Y., Li, H., & Nie, Y. (2023) · IEEE Joint International Information Technology and Artificial Intelligence Conference (ITAIC) · doi:10.1109/itaic58329.2023.10408789 · Cited by 2
Scheduled Curiosity-Deep Dyna-Q: Efficient Exploration for Dialog Policy Learning
Niu, X., Ito, A., & Nose, T. (2024) · IEEE Access · doi:10.1109/access.2024.3376418 · Cited by 3
“To overcome this problem, Bengio et al. [12] proposed curriculum learning (CL), which presents relatively easy or simple examples at an early stage, inspired by human learning habits.”
BackgroundGHCL: Gaussian heuristic curriculum learning for Brain CT report generation
Shen, Q., Shi, Y., Zhang, X., Ji, J., Liu, Y., & Xu, H. (2024) · Multimedia Systems · doi:10.1007/s00530-024-01266-3 · Cited by 6
“Curriculum Learning [28] has been widely adopted in the fields of natural language processing [29–32] and computer vision [33–35], which allows the model to progressively learn from simple samples to more complex ones during training [36].”
LEAD: Liberal Feature-based Distillation for Dense Retrieval
Sun, H., Liu, X., Gong, Y., Dong, A., Lu, J., Zhang, Y., Yang, L., Majumder, R., & Duan, N. (2024) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3616855.3635774 · Cited by 5
CLPSD: Detecting Ethereum Phishing Scams based on Curriculum Learning
Hou, W., Cui, B., Chen, Y., et al. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10448274 · Cited by 2
“Curriculum learning is a learning approach that emulates the way humans learn in educational settings, where training samples are arranged in order of difficulty and gradually presented to the model, starting from easier examples and progressing to more challenging ones [14].”
BackgroundImproving Kinyarwanda Speech Recognition Via Semi-Supervised Learning
Nzeyimana, A. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10447447 · Cited by 3
“Curriculum learning [11] has been shown to improve robustness in speech recognition [5].”
BackgroundBoosted verification using siamese neural network with DiffBlock
Liu, J., Liu, J., Jiang, R., Gu, B., Chen, Y., & Shen, C. (2024) · The Visual Computer · doi:10.1007/s00371-024-03318-1 · Cited by 2
“Curriculum learning (CL) [1] seeks to train the model from simple samples and gradually learn from hard samples and complex knowledge. index CL = randint ( 0 , index l ), rand ( 0 , 1 ) < 0 .”
Lee, H., Bae, K., Ha, S. J., et al. (2024) · IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) · doi:10.1109/wacv57701.2024.00667 · Cited by 2
“To stabilize the training process, we employ curriculum learning [2].”
MethodData-driven educational decision-making model for curriculum optimization
Mekić, E., Fetahović, I., Kuk, K. V., Popovic, B. M., & Čisar, P. (2024) · Elsevier eBooks · doi:10.1016/b978-0-443-18851-0.00002-0 · Cited by 5
Robust Federated Learning Based on Chained Self-Paced Learning
Jia, Y., Li, W., Yan, M., & Yan, B. (2023) · International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI) · doi:10.1109/cei60616.2023.10527834 · Cited by 2
“Self-paced learning is a variant of curriculum learning (CL) [17], which is a training strategy inspired by the learning process of humans and animals.”
BackgroundATELIER: Service Tailored and Limited-Trust Network Analytics Using Cooperative Learning
Milani, M., Bega, D., Gramaglia, M., Serrano, P., & Mannweiler, C. (2024) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2024.3401746 · Cited by 3
“Other approaches, like the one firstly proposed by Benigo et al. [36] try to overcome dataset limitations by gradually increasing the dataset difficulty during the training, e.g., introducing new data-classes in sequential steps.”
Method
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.