We consider the question of embedding nodes with similar local neighborhoods together in embedding space, commonly referred to as "role embeddings." We propose RAE, an unsupervised framework that learns role embeddings. It combines a within-node loss function and a graph neural network (GNN) architecture to place nodes with similar local neighborhoods close in embedding space. We also propose a faster way of generating negative examples called neighbor shuffling, which quickly creates negative examples directly within batches. These techniques can be easily combined with existing GNN methods to create unsupervised role embeddings at scale. We then explore role action embeddings, which summarize the non-structural features in a node's neighborhood, leading to better performance on node classification tasks. We find that the model architecture proposed here provides strong performance on both graph and node classification tasks, in some cases competitive with semi-supervised methods.
Recent studies have demonstrated the vulnerability of deep convolutional neural networks against adversarial examples. Inspired by the observation that the intrinsic dimension of image data is much smaller than its pixel space dimension and the vulnerability of neural networks grows with the input dimension, we propose to embed high-dimensional input images into a low-dimensional space to perform classification. However, arbitrarily projecting the input images to a low-dimensional space without regularization will not improve the robustness of deep neural networks. Leveraging optimal transport theory, we propose a new framework, Optimal Transport Classifier (OT-Classifier), and derive an objective that minimizes the discrepancy between the distribution of the true label and the distribution of the OT-Classifier output. Experimental results on several benchmark datasets show that, our proposed framework achieves state-of-the-art performance against strong adversarial attack methods.
Recent studies on multi-label image classification have focused on designing more complex architectures of deep neural networks such as the use of attention mechanisms and region proposal networks. Although performance gains have been reported, the backbone deep models of the proposed approaches and the evaluation metrics employed in different works vary, making it difficult to compare each fairly. Moreover, due to the lack of properly investigated baselines, the advantage introduced by the proposed techniques are often ambiguous. To address these issues, we make a thorough investigation of the mainstream deep convolutional neural network architectures for multi-label image classification and present a strong baseline. With the use of proper data augmentation techniques and model ensembles, the basic deep architectures can achieve better performance than many existing more complex ones on three benchmark datasets, providing great insight for the future studies on multi-label image classification.
Agriculture is vital for human survival and remains a major driver of several economies around the world; more so in underdeveloped and developing economies. With increasing demand for food and cash crops, due to a growing global population and the challenges posed by climate change, there is a pressing need to increase farm outputs while incurring minimal costs. Previous machine vision technologies developed for selective weeding have faced the challenge of reliable and accurate weed detection. We present approaches for plant seedlings classification with a dataset that contains 4,275 images of approximately 960 unique plants belonging to 12 species at several growth stages. We compare the performances of two traditional algorithms and a Convolutional Neural Network (CNN), a deep learning technique widely applied to image recognition, for this task. Our findings show that CNN-driven seedling classification applications when used in farming automation has the potential to optimize crop yield and improve productivity and efficiency when designed appropriately.
The objective learning formulation is essential for the success of convolutional neural networks. In this work, we analyse thoroughly the standard learning objective functions for multi-class classification CNNs: softmax regression (SR) for single-label scenario and logistic regression (LR) for multi-label scenario. Our analyses lead to an inspiration of exploiting LR for single-label classification learning, and then the disclosing of the negative class distraction problem in LR. To address this problem, we develop two novel LR based objective functions that not only generalise the conventional LR but importantly turn out to be competitive alternatives to SR in single label classification. Extensive comparative evaluations demonstrate the model learning advantages of the proposed LR functions over the commonly adopted SR in single-label coarse-grained object categorisation and cross-class fine-grained person instance identification tasks. We also show the performance superiority of our method on clothing attribute classification in comparison to the vanilla LR function.
This paper presents a novel feature set for shape-only leaf identification motivated by real-world, mobile deployment. The feature set includes basic shape features, as well as signal features extracted from local area integral invariants (LAIIs), similar to curvature maps, at multiple scales. The proposed methodology is evaluated on a number of publicly available leaf datasets with comparable results to existing methods which make use of colour and texture features in addition to shape. Over 90% classification accuracy is achieved on most datasets, with top-four accuracy for these datasets reaching over 98%. Rotation and scale invariance of the proposed features are demonstrated, along with an evaluation of the generalisability of the approach for generic shape matching.
Hyperspectral images (HSI) contain a wealth of information over hundreds of contiguous spectral bands, making it possible to classify materials through subtle spectral discrepancies. However, the classification of this rich spectral information is accompanied by the challenges like high dimensionality, singularity, limited training samples, lack of labeled data samples, heteroscedasticity and nonlinearity. To address these challenges, we propose a semi-supervised graph based dimensionality reduction method named `semi-supervised spatial spectral regularized manifold local scaling cut' (S3RMLSC). The underlying idea of the proposed method is to exploit the limited labeled information from both the spectral and spatial domains along with the abundant unlabeled samples to facilitate the classification task by retaining the original distribution of the data. In S3RMLSC, a hierarchical guided filter (HGF) is initially used to smoothen the pixels of the HSI data to preserve the spatial pixel consistency. This step is followed by the construction of linear patches from the nonlinear manifold by using the maximal linear patch (MLP) criterion. Then the inter-patch and intra-patch dissimilarity matrices are constructed in both spectral and spatial domains by regularized manifold local scaling cut (RMLSC) and neighboring pixel manifold local scaling cut (NPMLSC) respectively. Finally, we obtain the projection matrix by optimizing the updated semi-supervised spatial-spectral between-patch and total-patch dissimilarity. The effectiveness of the proposed DR algorithm is illustrated with publicly available real-world HSI datasets.
Manner of articulation detection using deep neural networks require a priori knowledge of the attribute discriminative features or the decent phoneme alignments. However generating an appropriate phoneme alignment is complex and its performance depends on the choice of optimal number of senones, Gaussians, etc. In the first part of our work, we exploit the manner of articulation detection using connectionist temporal classification (CTC) which doesn't need any phoneme alignment. Later we modify the state-of-the-art character based posteriors generated by CTC using the manner of articulation CTC detector. Beam search decoding is performed on the modified posteriors and it's impact on open source datasets such as AN4 and LibriSpeech is observed.
We present a two-layer fully connected neuromorphic system based on a thin-film transistor (TFT)-type NOR flash memory array with multiple postsynaptic (POST) neurons. Unsupervised online learning by spike-timing-dependent plasticity (STDP) on the binary MNIST handwritten datasets is implemented, and its recognition result is determined by measuring firing rate of POST neurons. Using a proposed learning scheme, we investigate the impact of the number of POST neurons in terms of recognition rate. In this neuromorphic system, lateral inhibition function and homeostatic property are exploited for competitive learning of multiple POST neurons. The simulation results demonstrate unsupervised online learning of the full black-and-white MNIST handwritten digits by STDP, which indicates the performance of pattern recognition and classification without preprocessing of input patterns.
A key feature of magnetic resonance (MR) imaging is its ability to manipulate how the intrinsic tissue parameters of the anatomy ultimately contribute to the contrast properties of the final, acquired image. This flexibility, however, can lead to substantial challenges for segmentation algorithms, particularly supervised methods. These methods require atlases or training data, which are composed of MR image and labeled image pairs. In most cases, the training data are obtained with a fixed acquisition protocol, leading to suboptimal performance when an input data set that requires segmentation has differing contrast properties. This drawback is increasingly significant with the recent movement towards multi-center research studies involving multiple scanners and acquisition protocols. In this work, we propose a new framework for supervised segmentation approaches that is robust to contrast differences between the training MR image and the input image. Our approach uses a generative simulation model within the segmentation process to compensate for the contrast differences. We allow the contrast of the MR image in the training data to vary by simulating a new contrast from the corresponding label image. The model parameters are optimized by a cost function measuring the consistency between the input MR image and its simulation based on a current estimate of the segmentation labels. We provide a proof of concept of this approach by combining a supervised classifier with a simple simulation model, and apply the resulting algorithm to synthetic images and actual MR images.
The skeleton based gesture recognition is gaining more popularity due to its wide possible applications. The key issues are how to extract discriminative features and how to design the classification model. In this paper, we first leverage a robust feature descriptor, path signature (PS), and propose three PS features to explicitly represent the spatial and temporal motion characteristics, i.e., spatial PS (S_PS), temporal PS (T_PS) and temporal spatial PS (T_S_PS). Considering the significance of fine hand movements in the gesture, we propose an "attention on hand" (AOH) principle to define joint pairs for the S_PS and select single joint for the T_PS. In addition, the dyadic method is employed to extract the T_PS and T_S_PS features that encode global and local temporal dynamics in the motion. Secondly, without the recurrent strategy, the classification model still faces challenges on temporal variation among different sequences. We propose a new temporal transformer module (TTM) that can match the sequence key frames by learning the temporal shifting parameter for each input. This is a learning-based module that can be included into standard neural network architecture. Finally, we design a multi-stream fully connected layer based network to treat spatial and temporal features separately and fused them together for the final result. We have tested our method on three benchmark gesture datasets, i.e., ChaLearn 2016, ChaLearn 2013 and MSRC-12. Experimental results demonstrate that we achieve the state-of-the-art performance on skeleton-based gesture recognition with high computational efficiency.
Transfer learning is a widely used method to build high performing computer vision models. In this paper, we study the efficacy of transfer learning by examining how the choice of data impacts performance. We find that more pre-training data does not always help, and transfer performance depends on a judicious choice of pre-training data. These findings are important given the continued increase in dataset sizes. We further propose domain adaptive transfer learning, a simple and effective pre-training method using importance weights computed based on the target dataset. Our method to compute importance weights follow from ideas in domain adaptation, and we show a novel application to transfer learning. Our methods achieve state-of-the-art results on multiple fine-grained classification datasets and are well-suited for use in practice.
Automated grading of prostate cancer histopathology images is a challenging task, with one key challenge being the scarcity of annotations down to the level of regions of interest (strong labels), as typically the prostate cancer Gleason score is known only for entire tissue slides (weak labels). In this study, we focus on automated Gleason score assignment of prostate cancer whole-slide images on the basis of a large weakly-labeled dataset and a smaller strongly-labeled one. We efficiently leverage information from both label sources by jointly training a classifier on the two datasets and by introducing a gradient update scheme that assigns different relative importances to each training example, as a means of self-controlling the weak supervision signal. Our approach achieves superior performance when compared with standard Gleason scoring methods.
Deep learning is extremely computationally intensive, and hardware vendors have responded by building faster accelerators in large clusters. Training deep learning models at petaFLOPS scale requires overcoming both algorithmic and systems software challenges. In this paper, we discuss three systems-related optimizations: (1) distributed batch normalization to control per-replica batch sizes, (2) input pipeline optimizations to sustain model throughput, and (3) 2-D torus all-reduce to speed up gradient summation. We combine these optimizations to train ResNet-50 on ImageNet to 76.3% accuracy in 2.2 minutes on a 1024-chip TPU v3 Pod with a training throughput of over 1.05 million images/second and no accuracy drop.
In assessing the severity of age-related macular degeneration (AMD), the Age-Related Eye Disease Study (AREDS) Simplified Severity Scale predicts the risk of progression to late AMD. However, its manual use requires the time-consuming participation of expert practitioners. Although several automated deep learning systems have been developed for classifying color fundus photographs (CFP) of individual eyes by AREDS severity score, none to date has used a patient-based scoring system that uses images from both eyes to assign a severity score. DeepSeeNet, a deep learning model, was developed to classify patients automatically by the AREDS Simplified Severity Scale (score 0-5) using bilateral CFP. DeepSeeNet was trained on 58,402 and tested on 900 images from the longitudinal follow-up of 4549 participants from AREDS. Gold standard labels were obtained using reading center grades. DeepSeeNet simulates the human grading process by first detecting individual AMD risk factors (drusen size, pigmentary abnormalities) for each eye and then calculating a patient-based AMD severity score using the AREDS Simplified Severity Scale. DeepSeeNet performed better on patient-based classification (accuracy = 0.671; kappa = 0.558) than retinal specialists (accuracy = 0.599; kappa = 0.467) with high AUC in the detection of large drusen (0.94), pigmentary abnormalities (0.93), and late AMD (0.97). DeepSeeNet demonstrated high accuracy with increased transparency in the automated assignment of individual patients to AMD risk categories based on the AREDS Simplified Severity Scale. These results highlight the potential of deep learning to assist and enhance clinical decision-making in patients with AMD, such as early AMD detection and risk prediction for developing late AMD. DeepSeeNet is publicly available on https://github.com/ncbi-nlp/DeepSeeNet.
Recent developments in gradient-based attention modeling have seen attention maps emerge as a powerful tool for interpreting convolutional neural networks. Despite good localization for an individual class of interest, these techniques produce attention maps with substantially overlapping responses among different classes, leading to the problem of visual confusion and the need for discriminative attention. In this paper, we address this problem by means of a new framework that makes class-discriminative attention a principled part of the learning process. Our key innovations include new learning objectives for attention separability and cross-layer consistency, which result in improved attention discriminability and reduced visual confusion. Extensive experiments on image classification benchmarks show the effectiveness of our approach in terms of improved classification accuracy, including CIFAR-100 (+3.33%), Caltech-256 (+1.64%), ILSVRC2012 (+0.92%), CUB-200-2011 (+4.8%) and PASCAL VOC2012 (+5.73%).
Although CNNs have gained the ability to transfer learned knowledge from source task to target task by virtue of large annotated datasets but consume huge processing time to fine-tune without GPU. In this paper, we propose a new computationally efficient transfer learning approach using classification layer features of pre-trained CNNs by appending layer after existing classification layer. We demonstrate that fine-tuning of the appended layer with existing classification layer for new task converges much faster than baseline and in average outperforms baseline classification accuracy. Furthermore, we execute thorough experiments to examine the influence of quantity, similarity, and dissimilarity of training sets in our classification outcomes to demonstrate transferability of classification layer features.
Visual object recognition plays an essential role in human daily life. This ability is so efficient that we can recognize a face or an object seemingly without effort, though they may vary in position, scale, pose, and illumination. In the field of computer vision, a large number of studies have been carried out to build a human-like object recognition system. Recently, deep neural networks have shown impressive progress in object classification performance, and have been reported to surpass humans. Yet there is still lack of thorough and fair comparison between humans and artificial recognition systems. While some studies consider artificially degraded images, human recognition performance on dataset widely used for deep neural networks has not been fully evaluated. The present paper carries out an extensive experiment to evaluate human classification accuracy on CIFAR10, a well-known dataset of natural images. This then allows for a fair comparison with the state-of-the-art deep neural networks. Our CIFAR10-based evaluations show very efficient object recognition of recent CNNs but, at the same time, prove that they are still far from human-level capability of generalization. Moreover, a detailed investigation using multiple levels of difficulty reveals that easy images for humans may not be easy for deep neural networks. Such images form a subset of CIFAR10 that can be employed to evaluate and improve future neural networks.
We consider a classifier whose test set is exposed to various perturbations that are not present in the training set. These test samples still contain enough features to map them to the same class as their unperturbed counterpart. Current architectures exhibit rapid degradation of accuracy when trained on standard datasets but then used to classify perturbed samples of that data. To address this, we present a novel architecture named DeepConsensus that significantly improves generalization to these test-time perturbations. Our key insight is that deep neural networks should directly consider summaries of low and high level features when making classifications. Existing convolutional neural networks can be augmented with DeepConsensus, leading to improved resistance against large and small perturbations on MNIST, EMNIST, FashionMNIST, CIFAR10 and SVHN datasets.
Although deep learning has shown great success in recent years, researchers have discovered a critical flaw where small, imperceptible changes in the input to the system can drastically change the output classification. These attacks are exploitable in nearly all of the existing deep learning classification frameworks. However, the susceptibility of deep sparse coding models to adversarial examples has not been examined. Here, we show that classifiers based on a deep sparse coding model whose classification accuracy is competitive with a variety of deep neural network models are robust to adversarial examples that effectively fool those same deep learning models. We demonstrate both quantitatively and qualitatively that the robustness of deep sparse coding models to adversarial examples arises from two key properties. First, because deep sparse coding models learn general features corresponding to generators of the dataset as a whole, rather than highly discriminative features for distinguishing specific classes, the resulting classifiers are less dependent on idiosyncratic features that might be more easily exploited. Second, because deep sparse coding models utilize fixed point attractor dynamics with top-down feedback, it is more difficult to find small changes to the input that drive the resulting representations out of the correct attractor basin.