Recent studies on learning with noisy labels have shown remarkable performance by exploiting a small clean dataset. In particular, model agnostic meta-learning-based label correction methods further improve performanc...
详细信息
ISBN:
(纸本)9783031198052;9783031198069
Recent studies on learning with noisy labels have shown remarkable performance by exploiting a small clean dataset. In particular, model agnostic meta-learning-based label correction methods further improve performance by correcting noisy labels on the fly. However, there is no safeguard on the label miscorrection, resulting in unavoidable performance degradation. Moreover, every training step requires at least three back-propagations, significantly slowing down the training speed. To mitigate these issues, we propose a robust and efficient method, FasTEN, which learns a label transitionmatrix on the fly. Employing the transitionmatrix makes the classifier skeptical about all the corrected samples, which alleviates the miscorrection issue. We also introduce a two-head architecture to efficiently estimate the label transitionmatrix every iteration within a single back-propagation, so that the estimated matrix closely follows the shifting noise distribution induced by label correction. Extensive experiments demonstrate that our FasTEN shows the best performance in training efficiency while having comparable or better accuracy than existing methods, especially achieving state-of-the-art performance in a real-world noisy dataset, Clothing1M.
In the realm of label noise learning, a potent strategy involves the application of noise transition matrices to foster robust learning processes. Most current research has gained significant success utilizing paramet...
详细信息
ISBN:
(纸本)9789819784981;9789819784998
In the realm of label noise learning, a potent strategy involves the application of noise transition matrices to foster robust learning processes. Most current research has gained significant success utilizing parameter estimation approaches to generate these matrices when facing instance-dependent noise. Nonetheless, a key drawback of this approach is the diffuse focus of the transitionmatrix, which can be indiscriminately distracting, thereby ignoring specific locations that are vulnerable to noise flipping. To address this gap, we introduce the Human Attention Constrained estimation (HACE). This innovative method capitalizes on human cognitive precedents to derive an inter-class affinity matrix. It further refines the estimation of the noise transitionmatrix by employing our novel matrix Structure Similarity (MSS) Loss, enabling the matrixestimation module to selectively concentrate on areas frequently affected by noisy flips. This targeted approach addresses the label noise conundrum more effectively and narrows the operational scope significantly. Experiments on three synthetic datasets and a real-world dataset corroborate the robustness and efficiency of our proposed method.
This study proposes a novel data-based approach for estimating the parameters of a stochastic hybrid model describing the traffic flow in an urban traffic network with signalized intersections. The model represents th...
详细信息
This study proposes a novel data-based approach for estimating the parameters of a stochastic hybrid model describing the traffic flow in an urban traffic network with signalized intersections. The model represents the evolution of the traffic flow rate, measuring the number of vehicles passing a given location per time unit. This traffic flow rate is described using a mode-dependent first-order autoregressive (AR) stochastic process. The parameters of the AR process take different values depending on the mode of traffic operation - free flowing, congested or faulty - making this a hybrid stochastic process. Mode switching occurs according to a first-order Markov chain. This study proposes an expectation-maximization (EM) technique for estimating the transitionmatrix of this Markovian mode process and the parameters of the AR models for each mode. The technique is applied to actual traffic flow data from the city of Jakarta, Indonesia. The model thus obtained is validated by using the smoothed inference algorithms and an online particle filter. The authors also develop an EM parameter estimation that, in combination with a time-window shift technique, can be useful and practical for periodically updating the parameters of hybrid model leading to an adaptive traffic flow state estimator.
暂无评论