As an intrinsic and fundamental property of big data, data heterogeneity exists in a variety of real-world applications, such as precision medicine, autonomous driving, financial applications, etc. For machine learning algorithms, the ignorance of data heterogeneity will greatly hurt the generalization performance and the algorithmic fairness, since the prediction mechanisms among different sub-populations are likely to differ from each other. In this work, we focus on the data heterogeneity that affects the prediction of machine learning models, and firstly propose the \emph{usable predictive heterogeneity}, which takes into account the model capacity and computational constraints. We prove that it can be reliably estimated from finite data with probably approximately correct (PAC) bounds. Additionally, we design a bi-level optimization algorithm to explore the usable predictive heterogeneity from data. Empirically, the explored heterogeneity provides insights for sub-population divisions in income prediction, crop yield prediction and image classification tasks, and leveraging such heterogeneity benefits the out-of-distribution generalization performance.
although best-in-class AI can deliver extraordinary outcomes in experimentation, data scientists struggle to duplicate these outcomes on actual-world data. It's nothing unexpected-actual data mirrors the messy world that made it, containing many biases and gaps. A painful element of real data is that it tends to be imbalanced. An imbalanced dataset is a dataset with a lot more examples in one class than others. This exploration features the broad study about taking care of class Imbalance issues utilizing Random Sampling and Data Augmentation Techniques. The critical angle featured is to grasp how Under-Sampling, Over-Sampling, and Data Augmentation use images and custom datasets. The model performance is improved with the expulsion of class imbalance issue utilizing different Augmentation approaches utilizing an augmentation library. The accuracy contrasts with taking care of the Class imbalance issue to boost accuracy, lessen error, and track down an ideal technique to tackle it.
The data in the real world consists of various kinds of painful features. A majorly found one is the class imbalance in which the number of examples in different classes in a dataset is unequal. The class imbalance is being resolved using various sampling techniques on the data. Augmentation technique Augmentation is one of the essential steps in any machine learning pipeline and is used for oversampling the data of minority classes. This paper aims to improve the model performance is being enhanced with removing the class imbalance problem by using various Augmentation approaches to generate various balanced augmented datasets using various data augmentation techniques & random sampling. The accuracies are acquired for each augmentation technique using a RESNET18 model. The model is run up to 100 epochs for each case, and the best accuracies are compared. This iterative comparison of various augmentation techniques has shown stunning insights into the effectiveness of multiple datasets.
Large Vision-Language Foundation Models (VLFM), such as CLIP, ALIGN and Florence, are trained on large-scale datasets of image-caption pairs and achieve superior transferability and robustness on downstream tasks, but they are difficult to use in many practical applications due to their large size, high latency and fixed architectures. Unfortunately, recent work shows training a small custom VLFM for resource-limited applications is currently very difficult using public and smaller-scale data. In this paper, we introduce a new distillation mechanism (DIME-FM) that allows us to transfer the knowledge contained in large VLFMs to smaller, customized foundation models using a relatively small amount of inexpensive, unpaired images and sentences. We transfer the knowledge from the pre-trained CLIP-ViTL/14 model to a ViT-B/32 model, with only 40M public images and 28.4M unpaired public sentences. The resulting model "Distill-ViT-B/32" rivals the CLIP-ViT-B/32 model pre-trained on its private WiT dataset (400M image-text pairs): Distill-ViT-B/32 achieves similar results in terms of zero-shot and linear-probing performance on both ImageNet and the ELEVATER (20 image classification tasks) benchmarks. It also displays comparable robustness when evaluated on five datasets with natural distribution shifts from ImageNet.
Classic Machine Learning techniques require training on data available in a single data lake. However, aggregating data from different owners is not always convenient for different reasons, including security, privacy and secrecy. Data carry a value that might vanish when shared with others; the ability to avoid sharing the data enables industrial applications where security and privacy are of paramount importance, making it possible to train global models by implementing only local policies which can be run independently and even on air-gapped data centres. Federated Learning (FL) is a distributed machine learning approach which has emerged as an effective way to address privacy concerns by only sharing local AI models while keeping the data decentralized. Two critical challenges of Federated Learning are managing the heterogeneous systems in the same federated network and dealing with real data, which are often not independently and identically distributed (non-IID) among the clients. In this paper, we focus on the second problem, i.e., the problem of statistical heterogeneity of the data in the same federated network. In this setting, local models might be strayed far from the local optimum of the complete dataset, thus possibly hindering the convergence of the federated model. Several Federated Learning algorithms, such as FedAvg, FedProx and Federated Curvature (FedCurv), aiming at tackling the non-IID setting, have already been proposed. This work provides an empirical assessment of the behaviour of FedAvg and FedCurv in common non-IID scenarios. Results show that the number of epochs per round is an important hyper-parameter that, when tuned appropriately, can lead to significant performance gains while reducing the communication cost. As a side product of this work, we release the non-IID version of the datasets we used so to facilitate further comparisons from the FL community.
Vision Transformers (ViTs) have been shown to be effective in various vision tasks. However, resizing them to a mobile-friendly size leads to significant performance degradation. Therefore, developing lightweight vision transformers has become a crucial area of research. This paper introduces CloFormer, a lightweight vision transformer that leverages context-aware local enhancement. CloFormer explores the relationship between globally shared weights often used in vanilla convolutional operators and token-specific context-aware weights appearing in attention, then proposes an effective and straightforward module to capture high-frequency local information. In CloFormer, we introduce AttnConv, a convolution operator in attention's style. The proposed AttnConv uses shared weights to aggregate local information and deploys carefully designed context-aware weights to enhance local features. The combination of the AttnConv and vanilla attention which uses pooling to reduce FLOPs in CloFormer enables the model to perceive high-frequency and low-frequency information. Extensive experiments were conducted in image classification, object detection, and semantic segmentation, demonstrating the superiority of CloFormer. The code is available at \url{https://github.com/qhfan/CloFormer}.
Vision Transformers (ViTs) have emerged as popular models in computer vision, demonstrating state-of-the-art performance across various tasks. This success typically follows a two-stage strategy involving pre-training on large-scale datasets using self-supervised signals, such as masked random patches, followed by fine-tuning on task-specific labeled datasets with cross-entropy loss. However, this reliance on cross-entropy loss has been identified as a limiting factor in ViTs, affecting their generalization and transferability to downstream tasks. Addressing this critical challenge, we introduce a novel Label-aware Contrastive Training framework, LaCViT, which significantly enhances the quality of embeddings in ViTs. LaCViT not only addresses the limitations of cross-entropy loss but also facilitates more effective transfer learning across diverse image classification tasks. Our comprehensive experiments on eight standard image classification datasets reveal that LaCViT statistically significantly enhances the performance of three evaluated ViTs by up-to 10.78% under Top-1 Accuracy.
Natural intelligences (NIs) thrive in a dynamic world - they learn quickly, sometimes with only a few samples. In contrast, artificial intelligences (AIs) typically learn with a prohibitive number of training samples and computational power. What design principle difference between NI and AI could contribute to such a discrepancy? Here, we investigate the role of weight polarity: development processes initialize NIs with advantageous polarity configurations; as NIs grow and learn, synapse magnitudes update, yet polarities are largely kept unchanged. We demonstrate with simulation and image classification tasks that if weight polarities are adequately set a priori, then networks learn with less time and data. We also explicitly illustrate situations in which a priori setting the weight polarities is disadvantageous for networks. Our work illustrates the value of weight polarities from the perspective of statistical and computational efficiency during learning.
Dense geometric matching determines the dense pixel-wise correspondence between a source and support image corresponding to the same 3D structure. Prior works employ an encoder of transformer blocks to correlate the two-frame features. However, existing monocular pretraining tasks, e.g., image classification, and masked image modeling (MIM), can not pretrain the cross-frame module, yielding less optimal performance. To resolve this, we reformulate the MIM from reconstructing a single masked image to reconstructing a pair of masked images, enabling the pretraining of transformer module. Additionally, we incorporate a decoder into pretraining for improved upsampling results. Further, to be robust to the textureless area, we propose a novel cross-frame global matching module (CFGM). Since the most textureless area is planar surfaces, we propose a homography loss to further regularize its learning. Combined together, we achieve the State-of-The-Art (SoTA) performance on geometric matching. Codes and models are available at https://github.com/ShngJZ/PMatch.
Contrastive learning methods train visual encoders by comparing views from one instance to others. Typically, the views created from one instance are set as positive, while views from other instances are negative. This binary instance discrimination is studied extensively to improve feature representations in self-supervised learning. In this paper, we rethink the instance discrimination framework and find the binary instance labeling insufficient to measure correlations between different samples. For an intuitive example, given a random image instance, there may exist other images in a mini-batch whose content meanings are the same (i.e., belonging to the same category) or partially related (i.e., belonging to a similar category). How to treat the images that correlate similarly to the current image instance leaves an unexplored problem. We thus propose to support the current image by exploring other correlated instances (i.e., soft neighbors). We first carefully cultivate a candidate neighbor set, which will be further utilized to explore the highly-correlated instances. A cross-attention module is then introduced to predict the correlation score (denoted as positiveness) of other correlated instances with respect to the current one. The positiveness score quantitatively measures the positive support from each correlated instance, and is encoded into the objective for pretext training. To this end, our proposed method benefits in discriminating uncorrelated instances while absorbing correlated instances for SSL. We evaluate our soft neighbor contrastive learning method (SNCLR) on standard visual recognition benchmarks, including image classification, object detection, and instance segmentation. The state-of-the-art recognition performance shows that SNCLR is effective in improving feature representations from both ViT and CNN encoders.
In this work, we present a data poisoning attack that confounds machine learning models without any manipulation of the image or label. This is achieved by simply leveraging the most confounding natural samples found within the training data itself, in a new form of a targeted attack coined "Mole Recruitment." We define moles as the training samples of a class that appear most similar to samples of another class, and show that simply restructuring training batches with an optimal number of moles can lead to significant degradation in the performance of the targeted class. We show the efficacy of this novel attack in an offline setting across several standard image classification datasets, and demonstrate the real-world viability of this attack in a continual learning (CL) setting. Our analysis reveals that state-of-the-art models are susceptible to Mole Recruitment, thereby exposing a previously undetected vulnerability of image classifiers.
Chest X-rays remains to be the most common imaging modality used to diagnose lung diseases. However, they necessitate the interpretation of experts (radiologists and pulmonologists), who are few. This review paper investigates the use of deep transfer learning techniques to detect COVID-19, pneumonia, and tuberculosis in chest X-ray (CXR) images. It provides an overview of current state-of-the-art CXR image classification techniques and discusses the challenges and opportunities in applying transfer learning to this domain. The paper provides a thorough examination of recent research studies that used deep transfer learning algorithms for COVID-19, pneumonia, and tuberculosis detection, highlighting the advantages and disadvantages of these approaches. Finally, the review paper discusses future research directions in the field of deep transfer learning for CXR image classification, as well as the potential for these techniques to aid in the diagnosis and treatment of lung diseases.
Distribution estimation has been demonstrated as one of the most effective approaches in dealing with few-shot image classification, as the low-level patterns and underlying representations can be easily transferred across different tasks in computer vision domain. However, directly applying this approach to few-shot text classification is challenging, since leveraging the statistics of known classes with sufficient samples to calibrate the distributions of novel classes may cause negative effects due to serious category difference in text domain. To alleviate this issue, we propose two simple yet effective strategies to estimate the distributions of the novel classes by utilizing unlabeled query samples, thus avoiding the potential negative transfer issue. Specifically, we first assume a class or sample follows the Gaussian distribution, and use the original support set and the nearest few query samples to estimate the corresponding mean and covariance. Then, we augment the labeled samples by sampling from the estimated distribution, which can provide sufficient supervision for training the classification model. Extensive experiments on eight few-shot text classification datasets show that the proposed method outperforms state-of-the-art baselines significantly.
The literature on provable robustness in machine learning has primarily focused on static prediction problems, such as image classification, in which input samples are assumed to be independent and model performance is measured as an expectation over the input distribution. Robustness certificates are derived for individual input instances with the assumption that the model is evaluated on each instance separately. However, in many deep learning applications such as online content recommendation and stock market analysis, models use historical data to make predictions. Robustness certificates based on the assumption of independent input samples are not directly applicable in such scenarios. In this work, we focus on the provable robustness of machine learning models in the context of data streams, where inputs are presented as a sequence of potentially correlated items. We derive robustness certificates for models that use a fixed-size sliding window over the input stream. Our guarantees hold for the average model performance across the entire stream and are independent of stream size, making them suitable for large data streams. We perform experiments on speech detection and human activity recognition tasks and show that our certificates can produce meaningful performance guarantees against adversarial perturbations.
In this paper, we study how the granularity of pretraining labels affects the generalization of deep neural networks in image classification tasks. We focus on the "fine-to-coarse" transfer learning setting, where the pretraining label space is more fine-grained than that of the target problem. Empirically, we show that pretraining on the leaf labels of ImageNet21k produces better transfer results on ImageNet1k than pretraining on other coarser granularity levels, which supports the common practice used in the community. Theoretically, we explain the benefit of fine-grained pretraining by proving that, for a data distribution satisfying certain hierarchy conditions, 1) coarse-grained pretraining only allows a neural network to learn the "common" or "easy-to-learn" features well, while 2) fine-grained pretraining helps the network learn the "rarer" or "fine-grained" features in addition to the common ones, thus improving its accuracy on hard downstream test samples in which common features are missing or weak in strength. Furthermore, we perform comprehensive experiments using the label hierarchies of iNaturalist 2021 and observe that the following conditions, in addition to proper choice of label granularity, enable the transfer to work well in practice: 1) the pretraining dataset needs to have a meaningful label hierarchy, and 2) the pretraining and target label functions need to align well.
It is broadly known that deep neural networks are susceptible to being fooled by adversarial examples with perturbations imperceptible by humans. Various defenses have been proposed to improve adversarial robustness, among which adversarial training methods are most effective. However, most of these methods treat the training samples independently and demand a tremendous amount of samples to train a robust network, while ignoring the latent structural information among these samples. In this work, we propose a novel Local Structure Preserving (LSP) regularization, which aims to preserve the local structure of the input space in the learned embedding space. In this manner, the attacking effect of adversarial samples lying in the vicinity of clean samples can be alleviated. We show strong empirical evidence that with or without adversarial training, our method consistently improves the performance of adversarial robustness on several image classification datasets compared to the baselines and some state-of-the-art approaches, thus providing promising direction for future research.
Deep learning models for image classification are typically trained under the "closed-world" assumption with a predefined set of image classes. However, when the models are deployed they may be faced with input images not belonging to the classes encountered during training. This type of scenario is common in remote sensing image classification where images come from different geographic areas, sensors, and imaging conditions. In this paper we deal with the problem of detecting remote sensing images coming from a different distribution compared to the training data - out of distribution images. We propose a benchmark for out of distribution detection in remote sensing scene classification and evaluate detectors based on maximum softmax probability and nearest neighbors. The experimental results show convincing advantages of the method based on nearest neighbors.
In this study, we address the emerging field of Streaming Federated Learning (SFL) and propose local cache update rules to manage dynamic data distributions and limited cache capacity. Traditional federated learning relies on fixed data sets, whereas in SFL, data is streamed, and its distribution changes over time, leading to discrepancies between the local training dataset and long-term distribution. To mitigate this problem, we propose three local cache update rules - First-In-First-Out (FIFO), Static Ratio Selective Replacement (SRSR), and Dynamic Ratio Selective Replacement (DRSR) - that update the local cache of each client while considering the limited cache capacity. Furthermore, we derive a convergence bound for our proposed SFL algorithm as a function of the distribution discrepancy between the long-term data distribution and the client's local training dataset. We then evaluate our proposed algorithm on two datasets: a network traffic classification dataset and an image classification dataset. Our experimental results demonstrate that our proposed local cache update rules significantly reduce the distribution discrepancy and outperform the baseline methods. Our study advances the field of SFL and provides practical cache management solutions in federated learning.
Diabetic Retinopathy (DR) causes quite a few blindness worldwide, which can be refrained by the timely diagnosis on retinal images. Recently, researches on deep learning-based retinal image classification have accelerated outstanding improvements in DR grading task. However, existing DR grading works are mostly limited to a supervised manner. They require accurately annotated data labeled by professional experts, and the annotating work is very laborious and time-consuming. We propose a Semi-supervised Auto-encoder Graph Network (SAGN) for the challenging DR diagnosis to relax this constraint. Precisely, SAGN consists of three major modules: auto-encoder feature learning, neighbor correlation mining, and graph representation. Firstly, our model learns to extract representations from retinal images and reconstruct them as close to original inputs as possible. Then neighbor correlations among labeled and unlabeled samples are established by their similarities, calculated by the radial basis function. Finally, we operate Graph Convolutional Neural Network (GCN) to grade retinal samples from extracted features and their correlations. To evaluate the performance of SAGN, we conduct sufficient comparative experiments on APTOS 2019 dataset, trained from EyePACS. Results demonstrate that our SAGN model can achieve comparable performance with limited labeled retinal images with the help of large amounts of unlabeled data.
Inspired by the long-range modeling ability of ViTs, large-kernel convolutions are widely studied and adopted recently to enlarge the receptive field and improve model performance, like the remarkable work ConvNeXt which employs 7x7 depthwise convolution. Although such depthwise operator only consumes a few FLOPs, it largely harms the model efficiency on powerful computing devices due to the high memory access costs. For example, ConvNeXt-T has similar FLOPs with ResNet-50 but only achieves ~60% throughputs when trained on A100 GPUs with full precision. Although reducing the kernel size of ConvNeXt can improve speed, it results in significant performance degradation, which poses a challenging problem: How to speed up large-kernel-based CNN models while preserving their performance. To tackle this issue, inspired by Inceptions, we propose to decompose large-kernel depthwise convolution into four parallel branches along channel dimension, i.e., small square kernel, two orthogonal band kernels, and an identity mapping. With this new Inception depthwise convolution, we build a series of networks, namely IncepitonNeXt, which not only enjoy high throughputs but also maintain competitive performance. For instance, InceptionNeXt-T achieves 1.6x higher training throughputs than ConvNeX-T, as well as attains 0.2% top-1 accuracy improvement on ImageNet-1K. We anticipate InceptionNeXt can serve as an economical baseline for future architecture design to reduce carbon footprint. Code is available at https://github.com/sail-sg/inceptionnext.