ICASSP 2023:ENTROPY BASED FEATURE REGULARIZATION TO IMPROVE TRANSFERABILITY OF DEEP LEARNING MODELS 分类任务中的标签往往只包含了数据集中的部分内容信息,例如自然图像中包含多个对象,但是标签中只有一个对象被标记。在使用 crossentropy 在这样的“粗标签”上进行训练时,对导致模型在“粗标签”上的过拟合,从而丢失...
RESUMO In this work, we describe a generalization of the standard MaxEnt regularization method. This generalization allows for a better exploitation of the available prior information about the expected structure of the true physical model, or of its derivatives, when solving...
terms are neglected.Instead, we propose a knowledge-free Entropy-based Attention Regularization (EAR) to discourage overfitting to training-specific terms. An additional objective function penalizes tokens with low self-attention entropy.We fine-tune BERT via EAR: the resulting model matches or exceeds...
To avoid overfitting, we establish a regularization term that is formulated as \({\mathscr {L}}_{reg} = \frac{1}{2}\sum _{i=0}^{I}(||W^i||_2^2 + ||\widehat{W^i}||^2_2)\), where \(W^i\) and \(\widehat{W^i}\) indicates the weight of the encoder and decoder ...
Dynamic clustering with complete fuzzy Informational Paradigm 2.1 LR Fuzzy data time array 2.2 Dissimilarity measures between LR fuzzy multivariate time trajectories 2.3 Fuzzy clustering models of LR fuzzy multivariate time trajectories based on entropy regularization 3. Application.University of Rome 'La ...
Infrared dim target detection based on total variation regularization and principal component pursuit 2017, Image and Vision Computing Show abstract Local Patch Network with Global Attention for Infrared Small Target Detection 2022, IEEE Transactions on Aerospace and Electronic Systems Single-Frame Infrared ...
Artifact-Free Wavelet Denoising: Non-convex Sparse Regularization, Convex Optimization. IEEE Signal Process. Lett. 2015, 22, 1364–1368. [Google Scholar] [CrossRef] Donoho, D.L. De-noising by soft thresholding. IEEE Trans. Inf. Theory 1995, 41, 613–627. [Google Scholar] [CrossRef] ...
Table 1 reports a comparison between the proposed method and two other clustering algorithms based on the incomplete Cholesky decomposition, namely algorithms (for a fair comparison, we report the results of the algorithm that does not use the 𝐿1L1 regularization) [18,22]. The comparison concer...
Symbolic Data Analysis provides suitable new types of variable that can take into account the variability present in the observed measurements. This paper proposes a partitioning fuzzy clustering algorithm for interval-valued data based on suitable adaptive Euclidean distance and entropy regularization. The...
A vector entropy consisting of the second order entropy (Ent-2) and the cross entropy is constructed as the regularization term which incorporates the prior motion knowledge into the estimation process. By imposing the motion constraints, the vector-entropy regularization converts the ill-posed ...