The pretasks are mainly built on mutual information estimation, which requires data augmentation to construct positive samples with similar semantics to learn invariant signals and negative samples with dissimilar semantics in order to empower representation discriminability. However, an appropriate data augmentation configuration depends heavily on lots of empirical trials such as choosing the compositions of data augmentation techniques and the corresponding hyperparameter settings. We propose an augmentation-free graph contrastive learning method, invariant-discriminative graph contrastive learning (iGCL), that does not intrinsically require negative samples. iGCL designs the invariant-discriminative loss (ID loss) to learn invariant and discriminative representations. On the one hand, ID loss learns invariant signals by directly minimizing the mean square error between the target samples and positive samples in the representation space. On the other hand, ID loss ensures that the representations are discriminative by an orthonormal constraint forcing the different dimensions of representations to be independent of each other. This prevents representations from collapsing to a point or subspace. Our theoretical analysis explains the effectiveness of ID loss from the perspectives of the redundancy reduction criterion, canonical correlation analysis, and information bottleneck principle. The experimental results demonstrate that iGCL outperforms all baselines on 5 node classification benchmark datasets. iGCL also shows superior performance for different label ratios and is capable of resisting graph attacks, which indicates that iGCL has excellent generalization and robustness.
Do we on the right way for remote sensing image understanding (RSIU) by training models via supervised data-dependent and task-dependent way, instead of human vision in a label-free and task-independent way? We argue that a more desirable RSIU model should be trained with intrinsic structure from data rather that extrinsic human labels to realize generalizability across a wide range of RSIU tasks. According to this hypothesis, we proposed \textbf{T}he \textbf{O}riginal \textbf{V}ision model (TOV) in remote sensing filed. Trained by massive unlabeled optical data along a human-like self-supervised learning (SSL) path that is from general knowledge to specialized knowledge, TOV model can be easily adapted to various RSIU tasks, including scene classification, object detection, and semantic segmentation, and outperforms dominant ImageNet supervised pretrained method as well as two recently proposed SSL pretrained methods on majority of 12 publicly available benchmarks. Moreover, we analyze the influences of two key factors on the performance of building TOV model for RSIU, including the influence of using different data sampling methods and the selection of learning paths during self-supervised optimization. We believe that a general model which is trained by a label-free and task-independent way may be the next paradigm for RSIU and hope the insights distilled from this study can help to foster the development of an original vision model for RSIU.
Point cloud scene flow estimation is of practical importance for dynamic scene navigation in autonomous driving. Since scene flow labels are hard to obtain, current methods train their models on synthetic data and transfer them to real scenes. However, large disparities between existing synthetic datasets and real scenes lead to poor model transfer. We make two major contributions to address that. First, we develop a point cloud collector and scene flow annotator for GTA-V engine to automatically obtain diverse realistic training samples without human intervention. With that, we develop a large-scale synthetic scene flow dataset GTA-SF. Second, we propose a mean-teacher-based domain adaptation framework that leverages self-generated pseudo-labels of the target domain. It also explicitly incorporates shape deformation regularization and surface correspondence refinement to address distortions and misalignments in domain transfer. Through extensive experiments, we show that our GTA-SF dataset leads to a consistent boost in model generalization to three real datasets (i.e., Waymo, Lyft and KITTI) as compared to the most widely used FT3D dataset. Moreover, our framework achieves superior adaptation performance on six source-target dataset pairs, remarkably closing the average domain gap by 60%. Data and codes are available at https://github.com/leolyj/DCA-SRSFE
Image segmentation is a crucial but challenging task that has many applications. In medical imaging for instance, intensity inhomogeneity and noise are common. In thigh muscle images, different muscles are closed packed together and there are often no clear boundaries between them. Intensity based segmentation models cannot separate one muscle from another. To solve such problems, in this work we present a segmentation model with adaptive spatial priors from joint registration. This model combines segmentation and registration in a unified framework to leverage their positive mutual influence. The segmentation is based on a modified Gaussian mixture model (GMM), which integrates intensity inhomogeneity and spacial smoothness. The registration plays the role of providing a shape prior. We adopt a modified sum of squared difference (SSD) fidelity term and Tikhonov regularity term for registration, and also utilize Gaussian pyramid and parametric method for robustness. The connection between segmentation and registration is guaranteed by the cross entropy metric that aims to make the segmentation map (from segmentation) and deformed atlas (from registration) as similar as possible. This joint framework is implemented within a constraint optimization framework, which leads to an efficient algorithm. We evaluate our proposed model on synthetic and thigh muscle MR images. Numerical results show the improvement as compared to segmentation and registration performed separately and other joint models.
Learning from a sequence of tasks for a lifetime is essential for an agent towards artificial general intelligence. This requires the agent to continuously learn and memorize new knowledge without interference. This paper first demonstrates a fundamental issue of lifelong learning using neural networks, named anterograde forgetting, i.e., preserving and transferring memory may inhibit the learning of new knowledge. This is attributed to the fact that the learning capacity of a neural network will be reduced as it keeps memorizing historical knowledge, and the fact that conceptual confusion may occur as it transfers irrelevant old knowledge to the current task. This work proposes a general framework named Cycled Memory Networks (CMN) to address the anterograde forgetting in neural networks for lifelong learning. The CMN consists of two individual memory networks to store short-term and long-term memories to avoid capacity shrinkage. A transfer cell is designed to connect these two memory networks, enabling knowledge transfer from the long-term memory network to the short-term memory network to mitigate the conceptual confusion, and a memory consolidation mechanism is developed to integrate short-term knowledge into the long-term memory network for knowledge accumulation. Experimental results demonstrate that the CMN can effectively address the anterograde forgetting on several task-related, task-conflict, class-incremental and cross-domain benchmarks.
Humans' continual learning (CL) ability is closely related to Stability Versus Plasticity Dilemma that describes how humans achieve ongoing learning capacity and preservation for learned information. The notion of CL has always been present in artificial intelligence (AI) since its births. This paper proposes a comprehensive review of CL. Different from previous reviews that mainly focus on the catastrophic forgetting phenomenon in CL, this paper surveys CL from a more macroscopic perspective based on the Stability Versus Plasticity mechanism. Analogous to biological counterpart, "smart" AI agents are supposed to i) remember previously learned information (information retrospection); ii) infer on new information continuously (information prospection:); iii) transfer useful information (information transfer), to achieve high-level CL. According to the taxonomy, evaluation metrics, algorithms, applications as well as some open issues are then introduced. Our main contributions concern i) rechecking CL from the level of artificial general intelligence; ii) providing a detailed and extensive overview on CL topics; iii) presenting some novel ideas on the potential development of CL.
Remembering and forgetting mechanisms are two sides of the same coin in a human learning-memory system. Inspired by human brain memory mechanisms, modern machine learning systems have been working to endow machine with lifelong learning capability through better remembering while pushing the forgetting as the antagonist to overcome. Nevertheless, this idea might only see the half picture. Up until very recently, increasing researchers argue that a brain is born to forget, i.e., forgetting is a natural and active process for abstract, rich, and flexible representations. This paper presents a learning model by active forgetting mechanism with artificial neural networks. The active forgetting mechanism (AFM) is introduced to a neural network via a "plug-and-play" forgetting layer (P\&PF), consisting of groups of inhibitory neurons with Internal Regulation Strategy (IRS) to adjust the extinction rate of themselves via lateral inhibition mechanism and External Regulation Strategy (ERS) to adjust the extinction rate of excitatory neurons via inhibition mechanism. Experimental studies have shown that the P\&PF offers surprising benefits: self-adaptive structure, strong generalization, long-term learning and memory, and robustness to data and parameter perturbation. This work sheds light on the importance of forgetting in the learning process and offers new perspectives to understand the underlying mechanisms of neural networks.
Graph neural networks (GNNs) have achieved great success in many graph-based tasks. Much work is dedicated to empowering GNNs with the adaptive locality ability, which enables measuring the importance of neighboring nodes to the target node by a node-specific mechanism. However, the current node-specific mechanisms are deficient in distinguishing the importance of nodes in the topology structure. We believe that the structural importance of neighboring nodes is closely related to their importance in aggregation. In this paper, we introduce discrete graph curvature (the Ricci curvature) to quantify the strength of structural connection of pairwise nodes. And we propose Curvature Graph Neural Network (CGNN), which effectively improves the adaptive locality ability of GNNs by leveraging the structural property of graph curvature. To improve the adaptability of curvature to various datasets, we explicitly transform curvature into the weights of neighboring nodes by the necessary Negative Curvature Processing Module and Curvature Normalization Module. Then, we conduct numerous experiments on various synthetic datasets and real-world datasets. The experimental results on synthetic datasets show that CGNN effectively exploits the topology structure information, and the performance is improved significantly. CGNN outperforms the baselines on 5 dense node classification benchmark datasets. This study deepens the understanding of how to utilize advanced topology information and assign the importance of neighboring nodes from the perspective of graph curvature and encourages us to bridge the gap between graph theory and neural networks.
A new learning paradigm, self-supervised learning (SSL), can be used to solve such problems by pre-training a general model with large unlabeled images and then fine-tuning on a downstream task with very few labeled samples. Contrastive learning is a typical method of SSL, which can learn general invariant features. However, most of the existing contrastive learning is designed for classification tasks to obtain an image-level representation, which may be sub-optimal for semantic segmentation tasks requiring pixel-level discrimination. Therefore, we propose Global style and Local matching Contrastive Learning Network (GLCNet) for remote sensing semantic segmentation. Specifically, the global style contrastive module is used to learn an image-level representation better, as we consider the style features can better represent the overall image features; The local features matching contrastive module is designed to learn representations of local regions which is beneficial for semantic segmentation. We evaluate four remote sensing semantic segmentation datasets, and the experimental results show that our method mostly outperforms state-of-the-art self-supervised methods and ImageNet pre-training. Specifically, with 1\% annotation from the original dataset, our approach improves Kappa by 6\% on the ISPRS Potsdam dataset and 3\% on Deep Globe Land Cover Classification dataset relative to the existing baseline. Moreover, our method outperforms supervised learning when there are some differences between the datasets of upstream tasks and downstream tasks. Our study promotes the development of self-supervised learning in the field of remote sensing semantic segmentation. The source code is available at https://github.com/GeoX-Lab/G-RSIM.
Considering the success of generative adversarial networks (GANs) for image-to-image translation, researchers have attempted to translate remote sensing images (RSIs) to maps (rs2map) through GAN for cartography. However, these studies involved limited scales, which hinders multi-scale map creation. By extending their method, multi-scale RSIs can be trivially translated to multi-scale maps (multi-scale rs2map translation) through scale-wise rs2map models trained for certain scales (parallel strategy). However, this strategy has two theoretical limitations. First, inconsistency between various spatial resolutions of multi-scale RSIs and object generalization on multi-scale maps (RS-m inconsistency) increasingly complicate the extraction of geographical information from RSIs for rs2map models with decreasing scale. Second, as rs2map translation is cross-domain, generators incur high computation costs to transform the RSI pixel distribution to that on maps. Thus, we designed a series strategy of generators for multi-scale rs2map translation to address these limitations. In this strategy, high-resolution RSIs are inputted to an rs2map model to output large-scale maps, which are translated to multi-scale maps through series multi-scale map translation models. The series strategy avoids RS-m inconsistency as inputs are high-resolution large-scale RSIs, and reduces the distribution gap in multi-scale map generation through similar pixel distributions among multi-scale maps. Our experimental results showed better quality multi-scale map generation with the series strategy, as shown by average increases of 11.69%, 53.78%, 55.42%, and 72.34% in the structural similarity index, edge structural similarity index, intersection over union (road), and intersection over union (water) for data from Mexico City and Tokyo at zoom level 17-13.