State-of-health (SOH) estimation is a crucial step for the operation of lithium-ion batteries (LIBs). Current SOH estimation methods suffer from various operating conditions and high measurement noise. A lightweight hierarchical temporal convolution network (LHTCNet) is proposed for estimating the SOH of LIBs. Specifically, hierarchical average pooling is used to draw multi-scale subsequences from the input data, and it is processed in parallel by a multi-level feature extraction module to capture local degradation and long term trend. A new multi-stage feature fusion module combines multi-scales via hierarchical concatenation to preserve critical degradation information in different granularities, and a estimate module outputs the final SOH estimate. Experiments conducted on two public data sets reveal the merits of LHTCNet. Also, the proposed model has very well generalized performance, maintaining high stability even when the noise-loss is very severe and partial charge-discharge conditions are severe. With only 121.57K parameters, the high computational efficiency of LHTCNet make it suitable for the resource-constrained edge devices in electric vehicles and energy storage devices.

Visual In-Context Learning (VICL) has emerged as a prominent approach for adapting visual foundation models to novel tasks, by effectively exploiting contextual information embedded in in-context examples, which can be formulated as a global ranking problem of potential candidates. Current VICL methods, such as Partial2Global and VPR, are grounded in the similarity-priority assumption that images more visually similar to a query image serve as better in-context examples. This foundational assumption, while intuitive, lacks sufficient justification for its efficacy in selecting optimal in-context examples. Furthermore, Partial2Global constructs its global ranking from a series of randomly sampled pairwise preference predictions. Such a reliance on random sampling can lead to incomplete coverage and redundant samplings of comparisons, thus further adversely impacting the final global ranking. To address these issues, this paper introduces an enhanced variant of Partial2Global designed for reliable and holistic selection of in-context examples in VICL. Our proposed method, dubbed RH-Partial2Global, leverages a jackknife conformal prediction-guided strategy to construct reliable alternative sets and a covering design-based sampling approach to ensure comprehensive and uniform coverage of pairwise preferences. Extensive experiments demonstrate that RH-Partial2Global achieves excellent performance and outperforms Partial2Global across diverse visual tasks.

In recent years, the field of image generation has witnessed significant advancements, particularly in fine-tuning methods that align models with universal human preferences. This paper explores the critical role of preference data in the training process of diffusion models, particularly in the context of Diffusion-DPO and its subsequent adaptations. We investigate the complexities surrounding universal human preferences in image generation, highlighting the subjective nature of these preferences and the challenges posed by minority samples in preference datasets. Through pilot experiments, we demonstrate the existence of minority samples and their detrimental effects on model performance. We propose Adaptive-DPO – a novel approach that incorporates a minority-instance-aware metric into the DPO objective. This metric, which includes intra-annotator confidence and inter-annotator stability, distinguishes between majority and minority samples. We introduce an Adaptive-DPO loss function which improves the DPO loss in two ways: enhancing the model's learning of majority labels while mitigating the negative impact of minority samples. Our experiments demonstrate that this method effectively handles both synthetic minority data and real-world preference data, paving the way for more effective training methodologies in image generation tasks.

Large language models have shown their ability to become effective few-shot learners with prompting, revoluting the paradigm of learning with data scarcity. However, this approach largely depends on the quality of prompt initialization and always exhibits large variability among different runs. Such property makes prompt tuning highly unreliable and vulnerable to poorly constructed prompts, which limits its extension to more real-world applications. To tackle this issue, we propose to treat the hard prompt and soft prompt as separate inputs to mitigate noise brought by the prompt initialization. Furthermore, we optimize soft prompts with contrastive learning for utilizing class-aware information in the training process to maintain model performance. Experimental results demonstrate that StablePT outperforms state-of-the-art methods by 6.97% in accuracy and reduces the standard deviation by 1.92 on average. Furthermore, extensive experiments underscore its robustness and stability across 8 datasets covering various tasks.

Cross-modality distillation arises as an important topic for data modalities containing limited knowledge such as depth maps and high-quality sketches. Such techniques are of great importance, especially for memory and privacy-restricted scenarios where labeled training data is generally unavailable. To solve the problem, existing label-free methods leverage a few pairwise unlabeled data to distill the knowledge by aligning features or statistics between the source and target modalities. For instance, one typically aims to minimize the L2 distance between the learned features of pairs of samples in the source (e.g. image) and the target (e.g. sketch) modalities. However, these approaches only consider the positive correspondence in paired samples, which is typically limited in quantity, while overlooking the potential information within the negative relationship present in the unpaired data, which is more abundant in cross-modality datasets. To exploit such a negative relationship which plays a vital role in learning discriminative feature representation, we propose a novel framework called generalizable cross-modality contrastive distillation (CMCD), built upon contrastive learning that leverages both positive and negative correspondence, towards a better distillation of generalizable features. Extensive experimental results show that our algorithm outperforms existing algorithms consistently by a margin of 2-3\% across diverse modalities and tasks, covering modalities of image, sketch, depth map, and audio and tasks of recognition and segmentation. Our convergence analysis reveals that the distance between source and target modalities significantly impacts the test error on downstream tasks within the target modality which is also validated by the empirical results.

Recent advances in prompt optimization have notably enhanced the performance of pre-trained language models (PLMs) on downstream tasks. However, the potential of optimized prompts on domain generalization has been under-explored. To explore the nature of prompt generalization on unknown domains, we conduct pilot experiments and find that (i) Prompts gaining more attention weight from PLMs' deep layers are more generalizable and (ii) Prompts with more stable attention distributions in PLMs' deep layers are more generalizable. Thus, we offer a fresh objective towards domain-generalizable prompts optimization named "Concentration", which represents the "lookback" attention from the current decoding token to the prompt tokens, to increase the attention strength on prompts and reduce the fluctuation of attention distribution. We adapt this new objective to popular soft prompt and hard prompt optimization methods, respectively. Extensive experiments demonstrate that our idea improves comparison prompt optimization methods by 1.42% for soft prompt generalization and 2.16% for hard prompt generalization in accuracy on the multi-source domain generalization setting, while maintaining satisfying in-domain performance. The promising results validate the effectiveness of our proposed prompt optimization objective and provide key insights into domain-generalizable prompts.

The great success of deep neural networks is built upon their over-parameterization, which smooths the optimization landscape without degrading the generalization ability. Despite the benefits of over-parameterization, a huge amount of parameters makes deep networks cumbersome in daily life applications. On the other hand, training neural networks without over-parameterization faces many practical problems, e.g., being trapped in the local optimal. Though techniques such as pruning and distillation are developed, they are expensive in fully training a dense network as backward selection methods; and there is still a void on systematically exploring forward selection methods for learning structural sparsity in deep networks. To fill in this gap, this paper proposes a new approach based on differential inclusions of inverse scale spaces. Specifically, our method can generate a family of models from simple to complex ones along the dynamics via coupling a pair of parameters, such that over-parameterized deep models and their structural sparsity can be explored simultaneously. This kind of differential inclusion scheme has a simple discretization, dubbed Deep structure splitting Linearized Bregman Iteration ( DessiLBI ), whose global convergence in learning deep networks could be established under the Kurdyka-Łojasiewicz framework. Particularly, we explore several applications of DessiLBI, including finding sparse structures of networks directly via the coupled structure parameter and growing networks from simple to complex ones progressively. Experimental evidence shows that our method achieves comparable and even better performance than the competitive optimizers in exploring the sparse structure of several widely used backbones on the benchmark datasets. Remarkably, with early stopping, our method unveils “winning tickets” in early epochs: the effective sparse network structures with comparable test accuracy to fully trained over-parameterized models, that are further transferable to similar alternative tasks. Furthermore, our method is able to grow networks efficiently with adaptive filter configurations, demonstrating the good performance with much less computational cost. Codes and models can be downloaded at https://github.com/DessiLBI2020/DessiLBI .

Positive-Unlabeled (PU) learning aims to learn a model with rare positive samples and abundant unlabeled samples. Compared with classical binary classification, the task of PU learning is much more challenging due to the existence of many incompletely-annotated data instances. Since only part of the most confident positive samples are available and evidence is not enough to categorize the rest samples, many of these unlabeled data may also be the positive samples. Research on this topic is particularly useful and essential to many real-world tasks which demand very expensive labelling cost. For example, the recognition tasks in disease diagnosis, recommendation system and satellite image recognition may only have few positive samples that can be annotated by the experts. These methods mainly omit the intrinsic hardness of some unlabeled data, which can result in sub-optimal performance as a consequence of fitting the easy noisy data and not sufficiently utilizing the hard data. In this paper, we focus on improving the commonly-used nnPU with a novel training pipeline. We highlight the intrinsic difference of hardness of samples in the dataset and the proper learning strategies for easy and hard data. By considering this fact, we propose first splitting the unlabeled dataset with an early-stop strategy. The samples that have inconsistent predictions between the temporary and base model are considered as hard samples. Then the model utilizes a noise-tolerant Jensen-Shannon divergence loss for easy data; and a dual-source consistency regularization for hard data which includes a cross-consistency between student and base model for low-level features and self-consistency for high-level features and predictions, respectively.

Dynamic Meta-filter Here we give the coding-level technical explanations of Eq. 2 and Eq. 3 in the main paper. Particularly, both equations consider operation on each position. (1) The procedure of Eq.2 is similar to normal convolution. A conv layer (kernel size 3, input channel c, output channel c*k*k) is applied to support feature, leading to output size (c*k*k)*w*h. The feature vector of position (i, j) with size c*k*k can then be written as in Eq.2. (2) For Eq.3, we first reshape fd(i, j) into size c*1*k*k, which is a convolution kernel with size k, group c, output channel c. It is applied to the corresponding query region, i.e., Bk(X :,i,j). The same procedure is applied to all (i, j). (3) The realization of Eq.3 follows idea of sliding window, where all windows with size k*k are extracted and convolved with corresponding DMF in form of inner production.

The research of generating images from scene graphs has become a hot topic, benefiting from the success of Generative Adversarial Network (GAN). Previous works in this field are still challenged by the complexity of texture pattern and object structure in images. In fact, it is more desirable to generate sketches directly, rather than image synthesis from scene graphs. The sketch is an abstract and iconic image representation, which describes the object structure and scene layout well but ignoring complex texture patterns, leading to a more reasonable generation task. Furthermore, there are two real-world tasks can illustrate the importance of this problem: Courtroom sketch and Crime Scene sketch, which are less studied before. To this end, we, for the first time, study the task of generating sketches from scene graphs. Essentially, the main novelties are two folds. First, a new sketch generation framework is developed which is trained with both newly designed perceptual loss and adversarial loss. Second, we propose a new layout encoding block to learn the depth information for each object. Extensive experiments on several widely used datasets validate the good performance of the proposed approach.

Few-shot learning (FSL) aims to recognize target classes by adapting the prior knowledge learned from source classes. Such knowledge usually resides in a deep embedding model for a general matching purpose of the support and query image pairs. The objective of this paper is to repurpose the contrastive learning for such matching to learn a few-shot embedding model. We make the following contributions: (i) We investigate the contrastive learning with Noise Contrastive Estimation (NCE) in a supervised manner for training a few-shot embedding model; (ii) We propose a novel contrastive training scheme dubbed infoPatch, exploiting the patch-wise relationship to substantially improve the popular infoNCE; (iii) We show that the embedding learned by the proposed infoPatch is more effective; (iv) Our model is thoroughly evaluated on few-shot recognition task; and demonstrates state-of-the-art results on miniImageNet and appealing performance on tieredImageNet, Fewshot-CIFAR100 (FC-100).

Few-shot learning (FSL), which aims to recognise new classes by adapting the learned knowledge with extremely limited few-shot (support) examples, remains an important open problem in computer vision. Most of the existing methods for feature alignment in few-shot learning only consider image-level or spatial-level alignment while omitting the channel disparity. Our insight is that these methods would lead to poor adaptation with redundant matching, and leveraging channel-wise adjustment is the key to well adapting the learned knowledge to new classes. Therefore, in this paper, we propose to learn a dynamic alignment, which can effectively highlight both query regions and channels according to different local support information. Specifically, this is achieved by first dynamically sampling the neighbourhood of the feature position conditioned on the input few shot, based on which we further predict a both position-dependent and channel-dependent Dynamic Meta-filter. The filter is used to align the query feature with position-specific and channel-specific knowledge. Moreover, we adopt Neural Ordinary Differential Equation (ODE) to enable a more accurate control of the alignment. In such a sense our model is able to better capture fine-grained semantic context of the few-shot example and thus facilitates dynamical knowledge adaptation for few-shot learning. The resulting framework establishes the new state-of-the-arts on major few-shot visual recognition benchmarks, including miniImageNet and tieredImageNet.

This paper studies the task of budgeted network learning [35] that aims at discovering good convolutional network structures under parameters/FLOPs constraints. Particularly, we present a novel Adaptive End-to-End Network Learning (AdeNeL) approach that enables learning the structures and parameters of networks simultaneously within the budgeted cost in terms of computation and memory consumption. We keep the depth of networks fixed to ensure a fair comparison with the backbones of competitors. Our AdeNeL learns to optimize both the parameters and the number of filters in each layer. To achieve this goal, our AdeNeL utilizes an iterative sparse regularization path – Dis-cretized Differential Inclusion of Inverse Scale Space (DI-ISS) to measure the capacity of the networks in the training process. Notably, the DI-ISS imports a group of augmented variables to explore the inverse scale space and this group of variables can be used to measure the redundancy of the network. According to the redundancy of the current network, our AdeNeL can choose appropriate operations for current network such as adding more filters. By this strategy, we can control the balance between the computational cost and the model performance in a dynamic way. Extensive experiments on several datasets including MNIST, CIFAR10/100, ImageNet with popular VGG and ResNet backbones validate the efficacy of our proposed method. In specific, on VGG16 backbone, our method on CIFAR10 achieves 92.71% test accuracy with only 0.58M (3.8%) parameters and 43M (13.7%) FLOPs, comparing to 92.90% test accuracy and 14.99 M parameters, 313M FLOPs of the full-sized VGG16 model.

In this paper, we propose an embarrassingly simple approach for one-shot learning. Our insight is that the one-shot tasks have domain gap to the network pretrained tasks and thus some features from the pretrained network are not relevant, or harmful to the specific one-shot task. Therefore, we propose to directly prune the features from the pretrained network for a specific one-shot task rather than update it via an optimized scheme with complex network structure. Without bells and whistles, our simple yet effective method achieves leading performances on miniImageNet (60.63%) and tieredImageNet (69.02%) for 5-way one-shot setting. The best trial can hit to 66.83% on miniImageNet and 74.04% on tieredImageNet, establishing a new state-of-the-art. We strongly advocate that our method can serve as a strong baseline for one-shot learning. The codes and trained models will be released at http://github.com/corwinliu9669/embarrassingly-simple-baseline.

Over-parameterization is ubiquitous nowadays in training neural networks to benefit both optimization in seeking global optima and generalization in reducing prediction error. However, compressive networks are desired in many real world applications and direct training of small networks may be trapped in local optima. In this paper, instead of pruning or distilling over-parameterized models to compressive ones, we propose a new approach based on \emph{differential inclusions of inverse scale spaces}, that generates a family of models from simple to complex ones by coupling gradient descent and mirror descent to explore model structural sparsity. It has a simple discretization, called the Split Linearized Bregman Iteration (SplitLBI), whose global convergence analysis in deep learning is established that from any initializations, algorithmic iterations converge to a critical point of empirical risks. Experimental evidence shows that\ SplitLBI may achieve state-of-the-art performance in large scale training on ImageNet-2012 dataset etc., while with \emph{early stopping} it unveils effective subnet architecture with comparable test accuracies to dense models after retraining instead of pruning well-trained ones.

Network structures are important to learning good representations of many tasks in computer vision and machine learning communities. These structures are either manually designed, or searched by Neural Architecture Search (NAS) in previous works, which however requires either expert-level efforts, or prohibitive computational cost. In practice, it is desirable to efficiently and simultaneously learn both the structures and parameters of a network from arbitrary classes with budgeted computational cost. We identify it as a new learning paradigm -- Boosting Network, where one starts from simple models, delving into complex trained models progressively. In this paper, by virtue of an iterative sparse regularization path -- Split Linearized Bregman Iteration (SplitLBI), we propose a simple yet effective boosting network method that can simultaneously grow and train a network by progressively adding both convolutional filters and layers. Extensive experiments with VGG and ResNets validate the effectiveness of our proposed algorithms.

Over-parameterization is ubiquitous nowadays in training neural networks to benefit both optimization in seeking global optima and generalization in reducing prediction error. However, compressive networks are desired in many real world applications and direct training of small networks may be trapped in local optima. In this paper, instead of pruning or distilling over-parameterized models to compressive ones, we propose a new approach based on differential inclusions of inverse scale spaces. Specifically, it generates a family of models from simple to complex ones that couples a pair of parameters to simultaneously train over-parameterized deep models and structural sparsity on weights of fully connected and convolutional layers. Such a differential inclusion scheme has a simple discretization, proposed as Deep structurally splitting Linearized Bregman Iteration (DessiLBI), whose global convergence analysis in deep learning is established that from any initializations, algorithmic iterations converge to a critical point of empirical risks. Experimental evidence shows that DessiLBI achieve comparable and even better performance than the competitive optimizers in exploring the structural sparsity of several widely used backbones on the benchmark datasets. Remarkably, with early stopping, DessiLBI unveils "winning tickets" in early epochs: the effective sparse structure with comparable test accuracy to fully trained over-parameterized models.

Over-parameterization is ubiquitous nowadays in training neural networks to benefit both optimization in seeking global optima and generalization in reducing prediction error. However, compressive networks are desired in many real world applications and direct training of small networks may be trapped in local optima. In this paper, instead of pruning or distilling an over-parameterized model to compressive ones, we propose a parsimonious learning approach based on differential inclusions of inverse scale spaces, that generates a family of models from simple to complex ones with a better efficiency and interpretability than stochastic gradient descent in exploring the model space. It enjoys a simple discretization, the Split Linearized Bregman Iterations, with provable global convergence that from any initializations, algorithmic iterations converge to a critical point of empirical risks. One may exploit the proposed method to boost the complexity of neural networks progressively. Numerical experiments with MNIST, Cifar-10/100, and ImageNet are conducted to show the method is promising in training large scale models with a favorite interpretability.

It becomes a research focus to build a Video-on-Demand (VOD) service which supports large-scale concurrent user access behavior. For its feature of multi-instance, packageable and high-performance, the virtual machine cluster will be a supporting technology to build large-scale VOD service. However, the existing virtual machine cluster scheduling method still have difficulty in dealing with the problem of random end-user behavior and hysteresis quality of virtual machine scheduling. A dynamic threshold-based pre-scheduling algorithm for VOD cluster in virtual environment was proposed. By predicting the VOD application workload when the bottleneck resource utilization ratio reaches the peak, the value of scheduling threshold was dynamically modified and then the virtual machine cluster was scheduled. The experiment results indicate that the dynamicthreshold-based per-scheduling algorithm can not only effectively reduce jitter incidence for VOD application, but also can improve the operating resource utilization ratio.