We address the challenging problem of whole slide image (WSI) classification. WSIs have very high resolutions and usually lack localized annotations. WSI classification can be cast as a multiple instance learning (MIL) problem when only slide-level labels are available. We propose a MIL-based method for WSI classification and tumor detection that does not require localized annotations. Our method has three major components. First, we introduce a novel MIL aggregator that models the relations of the instances in a dual-stream architecture with trainable distance measurement. Second, since WSIs can produce large or unbalanced bags that hinder the training of MIL models, we propose to use self-supervised contrastive learning to extract good representations for MIL and alleviate the issue of prohibitive memory cost for large bags. Third, we adopt a pyramidal fusion mechanism for multiscale WSI features, and further improve the accuracy of classification and localization. Our model is evaluated on two representative WSI datasets. The classification accuracy of our model compares favorably to fully-supervised methods, with less than 2% accuracy gap across datasets. Our results also outperform all previous MIL-based methods. Additional benchmark results on standard MIL datasets further demonstrate the superior performance of our MIL aggregator on general MIL problems. GitHub repository: https://github.com/binli123/dsmil-wsi
Although providing exceptional results for many computer vision tasks, state-of-the-art deep learning algorithms catastrophically struggle in low data scenarios. However, if data in additional modalities exist (e.g. text) this can compensate for the lack of data and improve the classification results. To overcome this data scarcity, we design a cross-modal feature generation framework capable of enriching the low populated embedding space in few-shot scenarios, leveraging data from the auxiliary modality. Specifically, we train a generative model that maps text data into the visual feature space to obtain more reliable prototypes. This allows to exploit data from additional modalities (e.g. text) during training while the ultimate task at test time remains classification with exclusively visual data. We show that in such cases nearest neighbor classification is a viable approach and outperform state-of-the-art single-modal and multimodal few-shot learning methods on the CUB-200 and Oxford-102 datasets.
Recent research has questioned the fairness of face-based recognition and attribute classification methods (such as gender and race) for dark-skinned people and women. Ocular biometrics in the visible spectrum is an alternate solution over face biometrics, thanks to its accuracy, security, robustness against facial expression, and ease of use in mobile devices. With the recent COVID-19 crisis, ocular biometrics has a further advantage over face biometrics in the presence of a mask. However, fairness of ocular biometrics has not been studied till now. This first study aims to explore the fairness of ocular-based authentication and gender classification methods across males and females. To this aim, VISOB $2.0$ dataset, along with its gender annotations, is used for the fairness analysis of ocular biometrics methods based on ResNet-50, MobileNet-V2 and lightCNN-29 models. Experimental results suggest the equivalent performance of males and females for ocular-based mobile user-authentication in terms of genuine match rate (GMR) at lower false match rates (FMRs) and an overall Area Under Curve (AUC). For instance, an AUC of 0.96 for females and 0.95 for males was obtained for lightCNN-29 on an average. However, males significantly outperformed females in deep learning based gender classification models based on ocular-region.
Explanation methods facilitate the development of models that learn meaningful concepts and avoid exploiting spurious correlations. We illustrate a previously unrecognized limitation of the popular neural network explanation method Grad-CAM: as a side effect of the gradient averaging step, Grad-CAM sometimes highlights locations the model did not actually use. To solve this problem, we propose HiResCAM, a novel class-specific explanation method that is guaranteed to highlight only the locations the model used to make each prediction. We prove that HiResCAM is a generalization of CAM and explore the relationships between HiResCAM and other gradient-based explanation methods. Experiments on PASCAL VOC 2012, including crowd-sourced evaluations, illustrate that while HiResCAM's explanations faithfully reflect the model, Grad-CAM often expands the attention to create bigger and smoother visualizations. Overall, this work advances convolutional neural network explanation approaches and may aid in the development of trustworthy models for sensitive applications.
We prove bounds for the approximation and estimation of certain binary classification functions using ReLU neural networks. Our estimation bounds provide a priori performance guarantees for empirical risk minimization using networks of a suitable size, depending on the number of training samples available. The obtained approximation and estimation rates are independent of the dimension of the input, showing that the curse of dimensionality can be overcome in this setting; in fact, the input dimension only enters in the form of a polynomial factor. Regarding the regularity of the target classification function, we assume the interfaces between the different classes to be locally of Barron-type. We complement our results by studying the relations between various Barron-type spaces that have been proposed in the literature. These spaces differ substantially more from each other than the current literature suggests.
Signet ring cell carcinoma (SRCC) of the stomach is a rare type of cancer with a slowly rising incidence. It tends to be more difficult to detect by pathologists mainly due to its cellular morphology and diffuse invasion manner, and it has poor prognosis when detected at an advanced stage. Computational pathology tools that can assist pathologists in detecting SRCC would be of a massive benefit. In this paper, we trained deep learning models using transfer learning, fully-supervised learning, and weakly-supervised learning to predict SRCC in Whole Slide Images (WSIs) using a training set of 1,765 WSIs. We evaluated the models on four different test sets of about 500 images each. The best model achieved a Receiver Operator Curve (ROC) area under the curve (AUC) of at least 0.99 on all four test sets, setting a top baseline performance for SRCC WSI classification.
The focal loss has demonstrated its effectiveness in many real-world applications such as object detection and image classification, but its theoretical understanding has been limited so far. In this paper, we first prove that the focal loss is classification-calibrated, i.e., its minimizer surely yields the Bayes-optimal classifier and thus the use of the focal loss in classification can be theoretically justified. However, we also prove a negative fact that the focal loss is not strictly proper, i.e., the confidence score of the classifier obtained by focal loss minimization does not match the true class-posterior probability and thus it is not reliable as a class-posterior probability estimator. To mitigate this problem, we next prove that a particular closed-form transformation of the confidence score allows us to recover the true class-posterior probability. Through experiments on benchmark datasets, we demonstrate that our proposed transformation significantly improves the accuracy of class-posterior probability estimation.
Deep neural networks generalize well despite being exceedingly overparameterized and being trained without explicit regularization. This curious phenomenon has inspired extensive research activity in establishing its statistical principles: Under what conditions is it observed? How do these depend on the data and on the training algorithm? When does regularization benefit generalization? While such questions remain wide open for deep neural nets, recent works have attempted gaining insights by studying simpler, often linear, models. Our paper contributes to this growing line of work by examining binary linear classification under a generative Gaussian mixture model. Motivated by recent results on the implicit bias of gradient descent, we study both max-margin SVM classifiers (corresponding to logistic loss) and min-norm interpolating classifiers (corresponding to least-squares loss). First, we leverage an idea introduced in [V. Muthukumar et al., arXiv:2005.08054, (2020)] to relate the SVM solution to the min-norm interpolating solution. Second, we derive novel non-asymptotic bounds on the classification error of the latter. Combining the two, we present novel sufficient conditions on the covariance spectrum and on the signal-to-noise ratio (SNR) under which interpolating estimators achieve asymptotically optimal performance as overparameterization increases. Interestingly, our results extend to a noisy model with constant probability noise flips. Contrary to previously studied discriminative data models, our results emphasize the crucial role of the SNR and its interplay with the data covariance. Finally, via a combination of analytical arguments and numerical demonstrations we identify conditions under which the interpolating estimator performs better than corresponding regularized estimates.
The softmax cross-entropy loss function has been widely used to train deep models for various tasks. In this work, we propose a Gaussian mixture (GM) loss function for deep neural networks for visual classification. Unlike the softmax cross-entropy loss, our method explicitly shapes the deep feature space towards a Gaussian Mixture distribution. With a classification margin and a likelihood regularization, the GM loss facilitates both high classification performance and accurate modeling of the feature distribution. The GM loss can be readily used to distinguish abnormal inputs, such as the adversarial examples, based on the discrepancy between feature distributions of the inputs and the training set. Furthermore, theoretical analysis shows that a symmetric feature space can be achieved by using the GM loss, which enables the models to perform robustly against adversarial attacks. The proposed model can be implemented easily and efficiently without using extra trainable parameters. Extensive evaluations demonstrate that the proposed method performs favorably not only on image classification but also on robust detection of adversarial examples generated by strong attacks under different threat models.
Whether what you see in Figure 1 is a "flamingo" or a "bird", is the question we ask in this paper. While fine-grained visual classification (FGVC) strives to arrive at the former, for the majority of us non-experts just "bird" would probably suffice. The real question is therefore -- how can we tailor for different fine-grained definitions under divergent levels of expertise. For that, we re-envisage the traditional setting of FGVC, from single-label classification, to that of top-down traversal of a pre-defined coarse-to-fine label hierarchy -- so that our answer becomes "bird"-->"Phoenicopteriformes"-->"Phoenicopteridae"-->"flamingo". To approach this new problem, we first conduct a comprehensive human study where we confirm that most participants prefer multi-granularity labels, regardless whether they consider themselves experts. We then discover the key intuition that: coarse-level label prediction exacerbates fine-grained feature learning, yet fine-level feature betters the learning of coarse-level classifier. This discovery enables us to design a very simple albeit surprisingly effective solution to our new problem, where we (i) leverage level-specific classification heads to disentangle coarse-level features with fine-grained ones, and (ii) allow finer-grained features to participate in coarser-grained label predictions, which in turn helps with better disentanglement. Experiments show that our method achieves superior performance in the new FGVC setting, and performs better than state-of-the-art on traditional single-label FGVC problem as well. Thanks to its simplicity, our method can be easily implemented on top of any existing FGVC frameworks and is parameter-free.
During a disaster event, images shared on social media helps crisis managers gain situational awareness and assess incurred damages, among other response tasks. Recent advances in computer vision and deep neural networks have enabled the development of models for real-time image classification for a number of tasks, including detecting crisis incidents, filtering irrelevant images, classifying images into specific humanitarian categories, and assessing the severity of damage. Despite several efforts, past works mainly suffer from limited resources (i.e., labeled images) available to train more robust deep learning models. In this study, we propose new datasets for disaster type detection, and informativeness classification, and damage severity assessment. Moreover, we relabel existing publicly available datasets for new tasks. We identify exact- and near-duplicates to form non-overlapping data splits, and finally consolidate them to create larger datasets. In our extensive experiments, we benchmark several state-of-the-art deep learning models and achieve promising results. We release our datasets and models publicly, aiming to provide proper baselines as well as to spur further research in the crisis informatics community.
We study the problem of predicting and controlling the future state distribution of an autonomous agent. This problem, which can be viewed as a reframing of goal-conditioned reinforcement learning (RL), is centered around learning a conditional probability density function over future states. Instead of directly estimating this density function, we indirectly estimate this density function by training a classifier to predict whether an observation comes from the future. Via Bayes' rule, predictions from our classifier can be transformed into predictions over future states. Importantly, an off-policy variant of our algorithm allows us to predict the future state distribution of a new policy, without collecting new experience. This variant allows us to optimize functionals of a policy's future state distribution, such as the density of reaching a particular goal state. While conceptually similar to Q-learning, our work lays a principled foundation for goal-conditioned RL as density estimation, providing justification for goal-conditioned methods used in prior work. This foundation makes hypotheses about Q-learning, including the optimal goal-sampling ratio, which we confirm experimentally. Moreover, our proposed method is competitive with prior goal-conditioned RL methods.
Neural image classification models typically consist of two components. The first is an image encoder, which is responsible for encoding a given raw image into a representative vector. The second is the classification component, which is often implemented by projecting the representative vector onto target class vectors. The target class vectors, along with the rest of the model parameters, are estimated so as to minimize the loss function. In this paper, we analyze how simple design choices for the classification layer affect the learning dynamics. We show that the standard cross-entropy training implicitly captures visual similarities between different classes, which might deteriorate accuracy or even prevents some models from converging. We propose to draw the class vectors randomly and set them as fixed during training, thus invalidating the visual similarities encoded in these vectors. We analyze the effects of keeping the class vectors fixed and show that it can increase the inter-class separability, intra-class compactness, and the overall model accuracy, while maintaining the robustness to image corruptions and the generalization of the learned concepts.
Identification of the type of communication technology and/or modulation scheme based on detected radio signal are challenging problems encountered in a variety of applications including spectrum allocation and radio interference mitigation. They are rendered difficult due to a growing number of emitter types and varied effects of real-world channels upon the radio signal. Existing spectrum monitoring techniques are capable of acquiring massive amounts of radio and real-time spectrum data using compact sensors deployed in a variety of settings. However, state-of-the-art methods that use such data to classify emitter types and detect communication schemes struggle to achieve required levels of accuracy at a computational efficiency that would allow their implementation on low-cost computational platforms. In this paper, we present a learning framework based on an LSTM denoising auto-encoder designed to automatically extract stable and robust features from noisy radio signals, and infer modulation or technology type using the learned features. The algorithm utilizes a compact neural network architecture readily implemented on a low-cost computational platform while exceeding state-of-the-art accuracy. Results on realistic synthetic as well as over-the-air radio data demonstrate that the proposed framework reliably and efficiently classifies received radio signals, often demonstrating superior performance compared to state-of-the-art methods.
Cardiovascular disease (CVD) is a serious illness affecting millions world-wide and is the leading cause of death in the US. Recent years, however, have seen tremendous growth in the area of personalized medicine, a field of medicine that places the patient at the center of the medical decision-making and treatment process. Many CVD-focused personalized medicine innovations focus on genetic biomarkers, which provide person-specific CVD insights at the genetic level, but do not focus on the practical steps a patient could take to mitigate their risk of CVD development. In this work we propose longitudinal inverse classification, a recommendation framework that provides personalized lifestyle recommendations that minimize the predicted probability of CVD risk. Our framework takes into account historical CVD risk, as well as other patient characteristics, to provide recommendations. Our experiments show that earlier adoption of the recommendations elicited from our framework produce significant CVD risk reduction.
The growth in the number of galaxy images is much faster than the speed at which these galaxies can be labelled by humans. However, by leveraging the information present in the ever growing set of unlabelled images, semi-supervised learning could be an effective way of reducing the required labelling and increasing classification accuracy. We develop a Variational Autoencoder (VAE) with Equivariant Transformer layers with a classifier network from the latent space. We show that this novel architecture leads to improvements in accuracy when used for the galaxy morphology classification task on the Galaxy Zoo data set. In addition we show that pre-training the classifier network as part of the VAE using the unlabelled data leads to higher accuracy with fewer labels compared to exiting approaches. This novel VAE has the potential to automate galaxy morphology classification with reduced human labelling efforts.
Quantifying uncertainty in a model's predictions is important as it enables the safety of an AI system to be increased by acting on the model's output in an informed manner. This is crucial for applications where the cost of an error is high, such as in autonomous vehicle control, medical image analysis, financial estimations or legal fields. Deep Neural Networks are powerful predictors that have recently achieved state-of-the-art performance on a wide spectrum of tasks. Quantifying predictive uncertainty in DNNs is a challenging and yet on-going problem. In this paper we propose a complete framework to capture and quantify three known types of uncertainty in DNNs for the task of image classification. This framework includes an ensemble of CNNs for model uncertainty, a supervised reconstruction auto-encoder to capture distributional uncertainty and using the output of activation functions in the last layer of the network, to capture data uncertainty. Finally we demonstrate the efficiency of our method on popular image datasets for classification.
The goal of semi-supervised learning is to utilize the unlabeled, in-domain dataset U to improve models trained on the labeled dataset D. Under the context of large-scale language-model (LM) pretraining, how we can make the best use of U is poorly understood: is semi-supervised learning still beneficial with the presence of large-scale pretraining? should U be used for in-domain LM pretraining or pseudo-label generation? how should the pseudo-label based semi-supervised model be actually implemented? how different semi-supervised strategies affect performances regarding D of different sizes, U of different sizes, etc. In this paper, we conduct comprehensive studies on semi-supervised learning in the task of text classification under the context of large-scale LM pretraining. Our studies shed important lights on the behavior of semi-supervised learning methods: (1) with the presence of in-domain pretraining LM on U, open-domain LM pretraining is unnecessary; (2) both the in-domain pretraining strategy and the pseudo-label based strategy introduce significant performance boosts, with the former performing better with larger U, the latter performing better with smaller U, and the combination leading to the largest performance boost; (3) self-training (pretraining first on pseudo labels D' and then fine-tuning on D) yields better performances when D is small, while joint training on the combination of pseudo labels D' and the original dataset D yields better performances when D is large. Using semi-supervised learning strategies, we are able to achieve a performance of around 93.8% accuracy with only 50 training data points on the IMDB dataset, and a competitive performance of 96.6% with the full IMDB dataset. Our work marks an initial step in understanding the behavior of semi-supervised learning models under the context of large-scale pretraining.
We study out-of-distribution (OOD) prediction behavior of neural networks when they classify images from unseen classes or corrupted images. To probe the OOD behavior, we introduce a new measure, nearest category generalization (NCG), where we compute the fraction of OOD inputs that are classified with the same label as their nearest neighbor in the training set. Our motivation stems from understanding the prediction patterns of adversarially robust networks, since previous work has identified unexpected consequences of training to be robust to norm-bounded perturbations. We find that robust networks have consistently higher NCG accuracy than natural training, even when the OOD data is much farther away than the robustness radius. This implies that the local regularization of robust training has a significant impact on the network's decision regions. We replicate our findings using many datasets, comparing new and existing training methods. Overall, adversarially robust networks resemble a nearest neighbor classifier when it comes to OOD data. Code available at https://github.com/yangarbiter/nearest-category-generalization.
Deep neural models have hitherto achieved significant performances on numerous classification tasks, but meanwhile require sufficient manually annotated data. Since it is extremely time-consuming and expensive to annotate adequate data for each classification task, learning an empirically effective model with generalization on small dataset has received increased attention. Existing efforts mainly focus on transferring task-relevant knowledge from other similar data to tackle the issue. These approaches have yielded remarkable improvements, yet neglecting the fact that the task-irrelevant features could bring out massive negative transfer effects. To date, no large-scale studies have been performed to investigate the impact of task-irrelevant features, let alone the utilization of this kind of features. In this paper, we firstly propose Task-Irrelevant Transfer Learning (TIRTL) to exploit task-irrelevant features, which mainly are extracted from task-irrelevant labels. Particularly, we suppress the expression of task-irrelevant information and facilitate the learning process of classification. We also provide a theoretical explanation of our method. In addition, TIRTL does not conflict with those that have previously exploited task-relevant knowledge and can be well combined to enable the simultaneous utilization of task-relevant and task-irrelevant features for the first time. In order to verify the effectiveness of our theory and method, we conduct extensive experiments on facial expression recognition and digit recognition tasks. Our source code will be also available in the future for reproducibility.