As an emerging and popular technique for boosting CNNs, structural reparameterization (SR) decouples the training and inference structures to alter the training dynamics and achieve cost-free ...improvement of a given network. Existing SR methods often prioritize network expressiveness enhancement but have yet to investigate approaches to mitigate significant bias and non-robustness of model prediction due to over-reliance on training data distribution and image noise. To this end, inspired by the effective strength of implicit regularization on the problem, this paper introduces an extra balanced implicit regularization mechanism into SR techniques to enhance the generalization of a given network for the first time. Specifically, we propose a novel SR module named DR-Block, which is used to complicate each convolutional layer of a given CNN during training. It draws on the advantages of deep matrix factorization with the regularization effect and further improves singular value dynamics by introducing batch normalization and dense connections to alleviate network degradation. At inference time, DR-Block can be equivalently reparameterized back into a single convolution for deployment. Furthermore, we empirically demonstrate the role of each design in DR-Block and explicitly reveal its inherent mechanism, which lies in enhancing the movement of large singular values while countering the attenuation of small ones. This helps enhance the interpretability of SR techniques. Experiments illustrate that DR-Block is an impressive alternative for a regular convolution layer of any structure and outperforms the existing SR methods in improving mainstream network architectures on various visual tasks. The code will be available online.
Latent Dirichlet allocation model (LDA) has been widely used in topic modeling. Recent works have shown the effectiveness of integrating neural network mechanisms with this generative model for ...learning text representation. However, one of the significant setbacks of LDA is that it is based on a Dirichlet prior that has a restrictive covariance structure. All its variables are considered to be negatively correlated, which makes the model restrictive. In a practical sense, topics can be positively or negatively correlated. To address this problem, we proposed a generalized Dirichlet variational autoencoder (GD-VAE) for topic modeling. The Generalized Dirichlet (GD) distribution has a more general covariance structure than the Dirichlet distribution because it takes into account both positively and negatively correlated topics in the corpus. Our proposed model leverages rejection sampling variational inference using a reparameterization trick for effective training. GD-VAE compares favorably to recent works on topic models on several benchmark corpora. Experiments show that accounting for topics’ positive and negative correlations results in better performance. We further validate the superiority of our proposed framework on two image data sets. GD-VAE demonstrates its significance as an integral part of a classification architecture. For reproducibility and further research purposes, code for this work can be found at https://github.com/hormone03/GD-VAE.
•We propose GD-VAE to capture correlations and learn complex distributions.•We show that capturing all correlations leads to improved performance in GD-VAE.•We address training instability by introducing a weighted objective function.•The comprehensive experiments show that GD-VAE outperformed state-of-the-art models.•We demonstrate the effectiveness of GD-VAE on data augmentation with image data sets.
Physics-based differentiable rendering is becoming increasingly crucial for tasks in inverse rendering and machine learning pipelines. To address discontinuities caused by geometric boundaries and ...occlusion, two classes of methods have been proposed: 1) the edge-sampling methods that directly sample light paths at the scene discontinuity boundaries, which require nontrivial data structures and precomputation to select the edges, and 2) the reparameterization methods that avoid discontinuity sampling but are currently limited to hemispherical integrals and unidirectional path tracing. We introduce a new mathematical formulation that enjoys the benefits of both classes of methods. Unlike previous reparameterization work that focused on hemispherical integral, we derive the reparameterization in the path space. As a result, to estimate derivatives using our formulation, we can apply advanced Monte Carlo rendering methods, such as bidirectional path tracing, while avoiding explicit sampling of discontinuity boundaries. We show differentiable rendering and inverse rendering results to demonstrate the effectiveness of our method.
•Modeling the uncertainty of attention modules.•Improving the generalization ability of attention models.•Mitigating the degradation issue that appears in the reparameterized attention.•Improving the ...image classification performance of different attention models on different datasets consistently.
The attention mechanism has been widely explored for neural networks as it could effectively model the interdependencies among channels, spatial positions, and frames. A neural network with attention modules has uncertainties in its parameters, but training the models deterministically hardly captures the uncertainties. Modeling the parameters’ uncertainty of the attention module could facilitate flexibly capturing the representative patterns, thus promoting the generalization of the models. In this work, we propose a novel reparameterized attention strategy by modeling the uncertainty of the parameters in the attention module and performing uncertainty-aware optimization. Instead of learning deterministic parameters for the attention modules, our strategy learns variational posterior distributions. The experimental results show that our strategy could consistently improve different models’ accuracy and reduce the generalization gap without extra computation.
In the context of medical image segmentation, the segmentation of polyps holds significant importance for early cancer screening and preoperative planning. Polyp images encapsulate substantial ...semantic information, including the main body of the colon, internal wall folds, and the polyp itself. These components exhibit numerous similar features, while the characteristics of polyps vary significantly in different locations. This paper proposes an improved U-Net neural network model designed to address the issue of unbalanced segmentation accuracy and generalization capability in existing models. The incorporation of a reparameterization module as the backbone network integrates multiscale features while adopting a training-prediction model separation pattern ensures the accuracy of the model. To avoid compromising global spatial information and simultaneously enhance the global perceptual capability, we employ a convolutional block attention module to compensate for the feature loss generated by skip connections. Additionally, we devise a loss computation method specific to this model, named CDLoss, to achieve more effective gradient optimization and enhance the model’s ability to segment polyp boundaries. Our model undergoes comprehensive validation on the Kavirs-SEG and CVC-ClinicDB datasets. The segmentation Intersection over Union (IoU) and Dice values reach 96.56% and 98.16%, respectively. The generalization capability achieves 98.57% and 97.33% in CVC-ClinicDB, surpassing other current state-of-the-art polyp image segmentation models.
•The reparameterization module helps balance the model’s extraction of detailed and broad features.•The skip-connected CBAM addresses semantic gaps in multiscale polyp features, enhancing RCNU-Net’s polyp segmentation with richer content.•The CDLoss function merges pixel-level classification accuracy and target overlap measurement to enhance model generalization.•Experimental results demonstrate the superiority of this model over previous research in polyp segmentation.
As urban traffic safety becomes increasingly important, real-time crosswalk detection is playing a critical role in the transportation field. However, existing crosswalk detection algorithms must be ...improved in terms of accuracy and speed. This study proposes a real-time crosswalk detector called X-CDNet based on YOLOX. Based on the ConvNeXt basic module, we designed a new basic module called Reparameterizable Sparse Large-Kernel (RepSLK) convolution that can be used to expand the model’s receptive field without the addition of extra inference time. In addition, we created a new crosswalk dataset called CD9K, which is based on realistic driving scenes augmented by techniques such as synthetic rain and fog. The experimental results demonstrate that X-CDNet outperforms YOLOX in terms of both detection accuracy and speed. X-CDNet achieves a 93.3 AP50 and a real-time detection speed of 123 FPS.
•Proposed a new basic module, RepSLK, for constructing backbone and neck networks.•Constructed a new real-time crosswalk detection model, X-CDNet.•Established a new crosswalk detection dataset, CD9K.
Sparse reparameterization in Deep Neural Networks (DNNs) aims to achieve a better tradeoff between the network parameter count and performance. Recently, the lottery ticket hypothesis suppose that ...excellent sub-networks (“winning tickets”) exist in dense randomly-initialized networks. These sparse sub-networks trained from scratch are able to reach the performance of their dense counterparts. Compared with Iterative Magnitude Pruning that relies on pruning strategies, the Continuous Sparsification algorithm learns the “winning tickets” with gradient-based methods, achieving better performance. In this paper, we propose Layer-wise Continuous Sparsification (LCS) scheme for finding sparse sub-networks, in which the parameterized relaxation of step functions used to remove network parameters in each layer is integrated into the DNN loss as an optimization objective. LCS utilizes a family of sigmoid functions to asynchronously filter important per-layer weights throughout training, yielding sparser and better sub-networks. Experiments show that our method surpasses state-of-the-art methods for sparse reparameterization. Additionally, the proposed method can be utilized as a regularization technique to further improve the accuracy of dense networks11Our code is publicly available at https://github.com/RiyaoDong/LCS..
In the domain of aquaculture, the act of feeding fish is a pivotal factor influencing both the growth of the fish and the associated cultivation costs. The implementation of intelligent feeding ...strategies is a crucial prerequisite for maintaining fish health and minimizing costs, with the accurate discernment of fish feeding behavior serving as the fundamental basis for the realization of such strategies. Addressing the issues of high data redundancy and substantial noise content inherent in the datasets utilized by existing identification models, as well as the intricate design and suboptimal execution efficiency of the model structures, this study introduces a two-stage framework for discriminating fish feeding behavior. In the initial stage, a re-parameterizable multi-scale object detection model is established, facilitating the acquisition of the spatial distribution of fish schools. This process effectively eliminates redundant data and noise, decouples the training and inference processes of the model, equivalently transforms the complex network training weights, and simplifies the inference process of the model. An analysis of the dataset characteristics is conducted, leading to the optimization of the model detector’s design and a reduction in both the model parameters and computational requirements. In the subsequent stage, responding to the key data characteristics of the spatial distribution of fish schools, a lightweight behavior recognition model is designed. This model enables the rapid and accurate identification of fish feeding behavior. A plethora of experimental results demonstrate that the proposed method can achieve a high recognition accuracy (Acc 83.33%) while operating under the constraints of minimal model parameters and computations (6.45M Params, 8.135G FLOPs). This provides a robust model foundation for the industrial application of the algorithm, underscoring the significant potential of the proposed method in advancing intelligent feeding strategies in aquaculture.
Display omitted
•A comprehensive dual-phase analytical framework is introduced for examining feeding patterns.•An innovative method integrating multi-scale information fusion and lightweight reparameterization backbone is proposed.•A classification model is presented to capture temporal and spatial complexities in fish feeding behaviors.•Achieved 83.33% accuracy, providing compelling evidence of its proficiency in analyzing fish feeding behavior.
A rational translational surface is a typical modeling surface used in computer-aided design and the architecture industry. In this study, we determine whether a given algebraic surface implicitly ...defined as V is a rational translational surface or not. This problem is reduced to finding the rational parameterizations of two space curves. More important, our discussions are constructive, and thus if V is translational, we provide a parametric representation of V of the form P(t1,t2)=P1(t1)+P2(t2).
Using Bayesian methods for extreme value analysis offers an alternative to frequentist ones, with several advantages such as easily dealing with parametric uncertainty or studying irregular models. ...However, computations can be challenging and the efficiency of algorithms can be altered by poor parametrization choices. The focus is on the Poisson process characterization of univariate extremes and outline two key benefits of an orthogonal parameterization. First, Markov chain Monte Carlo convergence is improved when applied on orthogonal parameters. This analysis relies on convergence diagnostics computed on several simulations. Second, orthogonalization also helps deriving Jeffreys and penalized complexity priors, and establishing posterior propriety thereof. The proposed framework is applied to return level estimation of Garonne flow data (France).
•An orthogonal parameterization is beneficial for Bayesian inference of extreme value models.•Convergence diagnostics (autocorrelation, ESS, and local Rˆ) are superior in the three maximum domains of attraction.•Jeffreys and PC priors can be computed for the Poisson process model for extremes, and the corresponding posterior is proper.•The return level credible interval length can be reduced by adding prior information on the extreme value index.