Problem Of Domain Generalization Research Articles

Sharpness-Aware Model-Agnostic Long-Tailed Domain Generalization

Domain Generalization (DG) aims to improve the generalization ability of models trained on a specific group of source domains, enabling them to perform well on new, unseen target domains. Recent studies have shown that methods that converge to smooth optima can enhance the generalization performance of supervised learning tasks such as classification. In this study, we examine the impact of smoothness-enhancing formulations on domain adversarial training, which combines task loss and adversarial loss objectives. Our approach leverages the fact that converging to a smooth minimum with respect to task loss can stabilize the task loss and lead to better performance on unseen domains. Furthermore, we recognize that the distribution of objects in the real world often follows a long-tailed class distribution, resulting in a mismatch between machine learning models and our expectations of their performance on all classes of datasets with long-tailed class distributions. To address this issue, we consider the domain generalization problem from the perspective of the long-tail distribution and propose using the maximum square loss to balance different classes which can improve model generalizability. Our method's effectiveness is demonstrated through comparisons with state-of-the-art methods on various domain generalization datasets. Code: https://github.com/bamboosir920/SAMALTDG.

Domain Generalization by Functional Regression

The problem of domain generalization is to learn, given data from different source distributions, a model that can be expected to generalize well on new target distributions which are only seen through unlabeled samples. In this paper, we study domain generalization as a problem of functional regression. Our concept leads to a new algorithm for learning a linear operator from marginal distributions of inputs to the corresponding conditional distributions of outputs given inputs. Our algorithm allows a source distribution-dependent construction of reproducing kernel Hilbert spaces for prediction, and, satisfies finite sample error bounds for the idealized risk. Numerical implementations and source code are available 1 .

Joint-product representation learning for domain generalization in classification and regression

In this work, we study the problem of generalizing a prediction (classification or regression) model trained on a set of source domains to an unseen target domain, where the source and target domains are different but related, i.e, the domain generalization problem. The challenge in this problem lies in the domain difference, which could degrade the generalization ability of the prediction model. To tackle this challenge, we propose to learn a neural network representation function to align a joint distribution and a product distribution in the representation space, and show that such joint-product distribution alignment conveniently leads to the alignment of multiple domains. In particular, we align the joint distribution and the product distribution under the L^{2}-distance, and show that this distance can be analytically estimated by exploiting its variational characterization and a linear variational function. This allows us to comfortably align the two distributions by minimizing the estimated distance with respect to the network representation function. Our experiments on synthetic and real-world datasets for classification and regression demonstrate the effectiveness of the proposed solution. For example, it achieves the best average classification accuracy of 82.26% on the text dataset Amazon Reviews, and the best average regression error of 0.114 on the WiFi dataset UJIIndoorLoc.

Research on Robust Audio-Visual Speech Recognition Algorithms

Automatic speech recognition (ASR) that relies on audio input suffers from significant degradation in noisy conditions and is particularly vulnerable to speech interference. However, video recordings of speech capture both visual and audio signals, providing a potent source of information for training speech models. Audiovisual speech recognition (AVSR) systems enhance the robustness of ASR by incorporating visual information from lip movements and associated sound production in addition to the auditory input. There are many audiovisual speech recognition models and systems for speech transcription, but most of them have been tested based in a single experimental setting and with a limited dataset. However, a good model should be applicable to any scenario. Our main contributions are: (i) Reproducing the three best-performing audiovisual speech recognition models in the current AVSR research area using the most famous audiovisual databases, LSR2 (Lip Reading Sentences 2) LSR3 (Lip Reading Sentences 3), and comparing and analyzing their performances under various noise conditions. (ii) Based on our experimental and research experiences, we analyzed the problems currently encountered in the AVSR domain, which are summarized as the feature-extraction problem and the domain-generalization problem. (iii) According to the experimental results, the Moco (momentum contrast) + word2vec (word to vector) model has the best AVSR effect on the LRS datasets regardless of whether there is noise or not. Additionally, the model also produced the best experimental results in the experiments of audio recognition and video recognition. Our research lays the foundation for further improving the performance of AVSR models.

Decomposed adversarial domain generalization

We tackle the problem of generalizing a predictor trained on a set of source domains to an unseen target domain, where the source and target domains are different but related to one another, i.e., the domain generalization problem. Prior adversarial methods rely on solving the minimax problems to align in the neural network embedding space the components of the domains (i.e., a set of marginal distributions, a set of marginal distributions and multiple sets of class-conditional distributions). However, these methods introduce additional parameters (for each set of distributions) to the network predictor and are difficult to train. In this work, we propose to directly align the domains themselves via solving a minimax problem that can be decomposed and converted into a min one. Particularly, we analytically solve the max problem with respect to (w.r.t.) the domain discriminators, and convert the minimax problem into a min one w.r.t. the embedding function. This is more advantageous since in the end our approach introduces no additional network parameters and simplifies the training procedure. We evaluate our approach on several multi-domain datasets and testify its superiority over the relevant methods. The source code is available at https://github.com/sentaochen/Decomposed-Adversarial-Domain-Generalization.

Contrastive-ACE : Domain Generalization Through Alignment of Causal Mechanisms.

Domain generalization aims to learn knowledge invariant across different distributions while semantically meaningful for downstream tasks from multiple source domains, to improve the model's generalization ability on unseen target domains. The fundamental objective is to understand the underlying "invariance" behind these observational distributions and such invariance has been shown to have a close connection to causality. While many existing approaches make use of the property that causal features are invariant across domains, we consider the invariance of the average causal effect of the features to the labels. This invariance regularizes our training approach in which interventions are performed on features to enforce stability of the causal prediction by the classifier across domains. Our work thus sheds some light on the domain generalization problem by introducing invariance of the mechanisms into the learning process. Experiments on several benchmark datasets demonstrate the performance of the proposed method against SOTAs.

Domain generalization in deep learning for contrast-enhanced imaging

Background:The domain generalization problem has been widely investigated in deep learning for non-contrast imaging over the last years, but it received limited attention for contrast-enhanced imaging. However, there are marked differences in contrast imaging protocols across clinical centers, in particular in the time between contrast injection and image acquisition, while access to multi-center contrast-enhanced image data is limited compared to available datasets for non-contrast imaging. This calls for new tools for generalizing single-domain, single-center deep learning models across new unseen domains and clinical centers in contrast-enhanced imaging. Methods:In this paper, we present an exhaustive evaluation of deep learning techniques to achieve generalizability to unseen clinical centers for contrast-enhanced image segmentation. To this end, several techniques are investigated, optimized and systematically evaluated, including data augmentation, domain mixing, transfer learning and domain adaptation. To demonstrate the potential of domain generalization for contrast-enhanced imaging, the methods are evaluated for ventricular segmentation in contrast-enhanced cardiac magnetic resonance imaging (MRI). Results:The results are obtained based on a multi-center cardiac contrast-enhanced MRI dataset acquired in four hospitals located in three countries (France, Spain and China). They show that the combination of data augmentation and transfer learning can lead to single-center models that generalize well to new clinical centers not included during training. Conclusions:Single-domain neural networks enriched with suitable generalization procedures can reach and even surpass the performance of multi-center, multi-vendor models in contrast-enhanced imaging, hence eliminating the need for comprehensive multi-center datasets to train generalizable models.

Unsupervised Domain Adaptation by Statistics Alignment for Deep Sleep Staging Networks.

Deep sleep staging networks have reached top performance on large-scale datasets. However, these models perform poorer when training and testing on small sleep cohorts due to data inefficiency. Transferring well-trained models from large-scale datasets (source domain) to small sleep cohorts (target domain) is a promising solution but still remains challenging due to the domain-shift issue. In this work, an unsupervised domain adaptation approach, domain statistics alignment (DSA), is developed to bridge the gap between the data distribution of source and target domains. DSA adapts the source models on the target domain by modulating the domain-specific statistics of deep features stored in the Batch Normalization (BN) layers. Furthermore, we have extended DSA by introducing cross-domain statistics in each BN layer to perform DSA adaptively (AdaDSA). The proposed methods merely need the well-trained source model without access to the source data, which may be proprietary and inaccessible. DSA and AdaDSA are universally applicable to various deep sleep staging networks that have BN layers. We have validated the proposed methods by extensive experiments on two state-of-the-art deep sleep staging networks, DeepSleepNet+ and U-time. The performance was evaluated by conducting various transfer tasks on six sleep databases, including two large-scale databases, MASS and SHHS, as the source domain, four small sleep databases as the target domain. Thereinto, clinical sleep records acquired in Huashan Hospital, Shanghai, were used. The results show that both DSA and AdaDSA could significantly improve the performance of source models on target domains, providing novel insights into the domain generalization problem in sleep staging tasks.

Planning Deficits in Children With Specific Language Impairment Are Reflected in Unnecessarily Awkward Grasps.

Specific language impairment (SLI) affects many children, but its symptomatology is still being characterized. An emerging view, which challenges the notion that SLI is specific to language, is that SLI may actually reflect a domain-general deficit in procedural learning. We explored an extension of this hypothesis that a core deficit in SLI involves a domain-general problem in planning. We used a dowel-transport task to study the extent to which 13 children with SLI and 14 typically developing (TD) controls (ages over both groups between 8;10 [years;months] and 12;11) would adopt initially awkward grasps that ensured comfortable final grasps when reaching out to move a dowel from 1 position to another (the end-state comfort effect). We predicted that children with SLI would be less likely to use end-state comfort grasps than would TD children. Contrary to our prediction, when awkward grasps were needed to ensure comfortable final grasps, participants with SLI showed the end-state comfort effect as often as the TD children did. Unexpectedly, however, in trials where awkward grasps were not needed for comfortable final grasps, the participants with SLI used more awkward grasps than did the TD participants after trials in which initial awkward grasps were needed. We suggest that this perseverative behavior is indicative of a domain-general problem in planning in SLI.

Domain-general problem solving skills and education in the 21st century

• Large-scale assessments such as PISA repeatedly emphasize the relevance of domain-general problem solving skills.

REEXAMINING THE FUNDAMENTAL DIFFERENCE HYPOTHESIS

The fundamental difference hypothesis (FDH), as formulated by Bley-Vroman (1990), claims that SLA tends to be nonconvergent because domain-specific linguistic mechanisms available in early childhood cannot be used for language learning in adulthood: Instead, second language (L2) learners deploy domain-general problem solving skills. I claim that nonconvergence is also true of some cases of unbalanced early bilingualism, when the target language was acquired in childhood, suggesting that nontarget attainment in these cases may have different roots—namely, inefficient learning mechanisms in L2 learners but insufficient input in early bilinguals. The FDH then predicts that early bilinguals should still be more successful at attaining nativelike knowledge than the L2 learners due to their early age of acquisition. This review article examines this prediction in light of recent studies of postpuberty L2 learners and unbalanced early bilinguals with nonconvergent knowledge of their first language. I show that the incidence of nativelike achievement is higher in the early bilingual groups than in the L2 groups and that differential performance by the two populations on different tasks lends some support to the processing claims of the FDH.