In recent years, many papers have reported state-of-the-art performance on Alzheimer's Disease classification with MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset using convolutional neural networks. However, we discover that when we split that data into training and testing sets at the subject level, we are not able to obtain similar performance, bringing the validity of many of the previous studies into question. Furthermore, we point out that previous works use different subsets of the ADNI data, making comparison across similar works tricky. In this study, we present the results of three splitting methods, discuss the motivations behind their validity, and report our results using all of the available subjects.
Typical video classification methods often divide a video into short clips, do inference on each clip independently, then aggregate the clip-level predictions to generate the video-level results. However, processing visually similar clips independently ignores the temporal structure of the video sequence, and increases the computational cost at inference time. In this paper, we propose a novel framework named FASTER, i.e., Feature Aggregation for Spatio-TEmporal Redundancy. FASTER aims to leverage the redundancy between neighboring clips and reduce the computational cost by learning to aggregate the predictions from models of different complexities. The FASTER framework can integrate high quality representations from expensive models to capture subtle motion information and lightweight representations from cheap models to cover scene changes in the video. A new recurrent network (i.e., FAST-GRU) is designed to aggregate the mixture of different representations. Compared with existing approaches, FASTER can reduce the FLOPs by over 10x? while maintaining the state-of-the-art accuracy across popular datasets, such as Kinetics, UCF-101 and HMDB-51.
Graph neural networks (GNNs) which apply the deep neural networks to graph data have achieved significant performance for the task of semi-supervised node classification. However, only few work has addressed the adversarial robustness of GNNs. In this paper, we first present a novel gradient-based attack method that facilitates the difficulty of tackling discrete graph data. When comparing to current adversarial attacks on GNNs, the results show that by only perturbing a small number of edge perturbations, including addition and deletion, our optimization-based attack can lead to a noticeable decrease in classification performance. Moreover, leveraging our gradient-based attack, we propose the first optimization-based adversarial training for GNNs. Our method yields higher robustness against both different gradient based and greedy attack methods without sacrificing classification accuracy on original graph.
This paper presents a novel method which simultaneously learns the number of filters and network features repeatedly over multiple epochs. We propose a novel pruning loss to explicitly enforces the optimizer to focus on promising candidate filters while suppressing contributions of less relevant ones. In the meanwhile, we further propose to enforce the diversities between filters and this diversity-based regularization term improves the trade-off between model sizes and accuracies. It turns out the interplay between architecture and feature optimizations improves the final compressed models, and the proposed method is compared favorably to existing methods, in terms of both models sizes and accuracies for a wide range of applications including image classification, image compression and audio classification.
Aspect-level sentiment classification aims to distinguish the sentiment polarities over one or more aspect terms in a sentence. Existing approaches mostly model different aspects in one sentence independently, which ignore the sentiment dependencies between different aspects. However, we find such dependency information between different aspects can bring additional valuable information. In this paper, we propose a novel aspect-level sentiment classification model based on graph convolutional networks (GCN) which can effectively capture the sentiment dependencies between multi-aspects in one sentence. Our model firstly introduces bidirectional attention mechanism with position encoding to model aspect-specific representations between each aspect and its context words, then employs GCN over the attention mechanism to capture the sentiment dependencies between different aspects in one sentence. We evaluate the proposed approach on the SemEval 2014 datasets. Experiments show that our model outperforms the state-of-the-art methods. We also conduct experiments to evaluate the effectiveness of GCN module, which indicates that the dependencies between different aspects is highly helpful in aspect-level sentiment classification.
Citation matching is a challenging task due to different problems such as the variety of citation styles, mistakes in reference strings and the quality of identified reference segments. The classic citation matching configuration used in this paper is the combination of blocking technique and a binary classifier. Three different possible inputs (reference strings, reference segments and a combination of reference strings and segments) were tested to find the most efficient strategy for citation matching. In the classification step, we describe the effect which the probabilities of reference segments can have in citation matching. Our evaluation on a manually curated gold standard showed that the input data consisting of the combination of reference segments and reference strings lead to the best result. In addition, the usage of the probabilities of the segmentation slightly improves the result.
This paper investigates deep neural networks for radio signal classification. Instead of performing modulation recognition and combining it with further analysis methods, the classifier operates directly on the IQ data of the signals and outputs the transmission mode. A data set of radio signals of 18 different modes, that commonly occur in the HF radio band, is presented and used as a showcase example. The data set considers HF channel properties and is used to train four different deep neural network architectures. The results of the best networks show an excellent accuracy of up to 98%.
With the widespread success of deep neural networks in science and technology, it is becoming increasingly important to quantify the uncertainty of the predictions produced by deep learning. In this paper, we introduce a new method that attaches an explicit uncertainty statement to the probabilities of classification using deep neural networks. Precisely, we view that the classification probabilities are sampled from an unknown distribution, and we propose to learn this distribution through the Dirichlet mixture that is flexible enough for approximating any continuous distribution on the simplex. We then construct credible intervals from the learned distribution to assess the uncertainty of the classification probabilities. Our approach is easy to implement, computationally efficient, and can be coupled with any deep neural network architecture. Our method leverages the crucial observation that, in many classification applications such as medical diagnosis, more than one class labels are available for each observational unit. We demonstrate the usefulness of our approach through simulations and a real data example.
Persistent homology analysis provides means to capture the connectivity structure of data sets in various dimensions. On the mathematical level, by defining a metric between the objects that persistence attaches to data sets, we can stabilize invariants characterizing these objects. We outline how so called contour functions induce relevant metrics for stabilizing the rank invariant. On the practical level, the stable ranks are used as fingerprints for data. Different choices of contour lead to different stable ranks and the topological learning is then the question of finding the optimal contour. We outline our analysis pipeline and show how it can enhance classification of physical activities data. As our main application we study how stable ranks and contours provide robust descriptors of spatial patterns of atmospheric cloud fields.
In this paper we address the problem of protein classification starting from a multi-view 2D representation of proteins. From each 3D protein structure, a large set of 2D projections is generated using the protein visualization software Jmol. This set of multi-view 2D representations includes 13 different types of protein visualizations that emphasize specific properties of protein structure (e.g., a backbone visualization that displays the backbone structure of the protein as a trace of the C{\alpha} atom). Each type of representation is used to train a different Convolutional Neural Network (CNN), and the fusion of these CNNs is shown to be able to exploit the diversity of different types of representations to improve classification performance. In addition, several multi-view projections are obtained by uniformly rotating the protein structure around its central X, Y, and Z viewing axes to produce 125 images. This approach can be considered a data augmentation method for improving the performance of the classifier and can be used in both the training and the testing phases. Experimental evaluation of the proposed approach on two datasets demonstrates the strength of the proposed method with respect to the other state-of-the-art approaches. The MATLAB code used in this paper is available at https://github.com/LorisNanni.
The analysis of electroencephalogram (EEG) waves is of critical importance for the diagnosis of sleep disorders, such as sleep apnea and insomnia, besides that, seizures, epilepsy, head injuries, dizziness, headaches and brain tumors. In this context, one important task is the identification of visible structures in the EEG signal, such as sleep spindles and K-complexes. The identification of these structures is usually performed by visual inspection from human experts, a process that can be error prone and susceptible to biases. Therefore there is interest in developing technologies for the automated analysis of EEG. In this paper, we propose a new Genetic Programming (GP) framework for feature construction and dimensionality reduction from EEG signals. We use these features to automatically identify spindles and K-complexes on data from the DREAMS project. Using 5 different classifiers, the set of attributes produced by GP obtained better AUC scores than those obtained from PCA or the full set of attributes. Also, the results obtained from the proposed framework obtained a better balance of Specificity and Recall than other models recently proposed in the literature. Analysis of the features most used by GP also suggested improvements for data acquisition protocols in future EEG examinations.
We develop new representations and algorithms for three-dimensional (3D) object detection and spatial layout prediction in cluttered indoor scenes. We first propose a clouds of oriented gradient (COG) descriptor that links the 2D appearance and 3D pose of object categories, and thus accurately models how perspective projection affects perceived image gradients. To better represent the 3D visual styles of large objects and provide contextual cues to improve the detection of small objects, we introduce latent support surfaces. We then propose a "Manhattan voxel" representation which better captures the 3D room layout geometry of common indoor environments. Effective classification rules are learned via a latent structured prediction framework. Contextual relationships among categories and layout are captured via a cascade of classifiers, leading to holistic scene hypotheses that exceed the state-of-the-art on the SUN RGB-D database.
Pruning unimportant parameters can allow deep neural networks (DNNs) to reduce their heavy computation and memory requirements. A saliency metric estimates which parameters can be safely pruned with little impact on the classification performance of the DNN. Many saliency metrics have been proposed, each within the context of a wider pruning algorithm. The result is that it is difficult to separate the effectiveness of the saliency metric from the wider pruning algorithm that surrounds it. Similar-looking saliency metrics can yield very different results because of apparently minor design choices. We propose a taxonomy of saliency metrics based on four mostly-orthogonal principal components. We show that a broad range of metrics from the pruning literature can be grouped according to these components. Our taxonomy not only serves as a guide to prior work, but allows us to construct new saliency metrics by exploring novel combinations of our taxonomic components. We perform an in-depth experimental investigation of more than 300 saliency metrics. Our results provide decisive answers to open research questions, and demonstrate the importance of reduction and scaling when pruning groups of weights. We find that some of our constructed metrics can outperform the best existing state-of-the-art metrics for convolutional neural network channel pruning.
We propose a novel paradigm for solving Inductive Logic Programming (ILP) problems via deep recurrent neural networks. This proposed ILP solver is designed based on differentiable implementation of the deduction via forward chaining. In contrast to the majority of past methods, instead of searching through the space of possible first-order logic rules by using some restrictive rule templates, we directly learn the symbolic logical predicate rules by introducing a novel differentiable Neural Logic (dNL) network. The proposed dNL network is able to learn and represent Boolean functions efficiently and in an explicit manner. We show that the proposed dNL-ILP solver supports desirable features such as recursion and predicate invention. Further, we investigate the performance of the proposed ILP solver in classification tasks involving benchmark relational datasets. In particular, we show that our proposed method outperforms the state of the art ILP solvers in classification tasks for Mutagenesis, Cora and IMDB datasets.
Reducing the latency variance in machine learning inference is a key requirement in many applications. Variance is harder to control in a cloud deployment in the presence of stragglers. In spite of this challenge, inference is increasingly being done in the cloud, due to the advent of affordable machine learning as a service (MLaaS) platforms. Existing approaches to reduce variance rely on replication which is expensive and partially negates the affordability of MLaaS. In this work, we argue that MLaaS platforms also provide unique opportunities to cut the cost of redundancy. In MLaaS platforms, multiple inference requests are concurrently received by a load balancer which can then create a more cost-efficient redundancy coding across a larger collection of images. We propose a novel convolutional neural network model, Collage-CNN, to provide a low-cost redundancy framework. A Collage-CNN model takes a collage formed by combining multiple images and performs multi-image classification in one shot, albeit at slightly lower accuracy. We then augment a collection of traditional single image classifiers with a single Collage-CNN classifier which acts as a low-cost redundant backup. Collage-CNN then provides backup classification results if a single image classification straggles. Deploying the Collage-CNN models in the cloud, we demonstrate that the 99th percentile tail latency of inference can be reduced by 1.47X compared to replication based approaches while providing high accuracy. Also, variation in inference latency can be reduced by 9X with a slight increase in average inference latency.
While paragraph embedding models are remarkably effective for downstream classification tasks, what they learn and encode into a single vector remains opaque. In this paper, we investigate a state-of-the-art paragraph embedding method proposed by Zhang et al. (2017) and discover that it cannot reliably tell whether a given sentence occurs in the input paragraph or not. We formulate a sentence content task to probe for this basic linguistic property and find that even a much simpler bag-of-words method has no trouble solving it. This result motivates us to replace the reconstruction-based objective of Zhang et al. (2017) with our sentence content probe objective in a semi-supervised setting. Despite its simplicity, our objective improves over paragraph reconstruction in terms of (1) downstream classification accuracies on benchmark datasets, (2) faster training, and (3) better generalization ability.
Exploiting rich spatial and spectral features contributes to improve the classification accuracy of hyperspectral images (HSIs). In this paper, based on the mechanism of the population receptive field (pRF) in human visual cortex, we further utilize the spatial correlation of pixels in images and propose pixel directed acyclic graph recurrent neural network (Pixel DAG-RNN) to extract and apply spectral-spatial features for HSIs classification. In our model, an undirected cyclic graph (UCG) is used to represent the relevance connectivity of pixels in an image patch, and four DAGs are used to approximate the spatial relationship of UCGs. In order to avoid overfitting, weight sharing and dropout are adopted. The higher classification performance of our model on HSIs classification has been verified by experiments on three benchmark data sets.
Polarimetric synthetic aperture radar (PolSAR) images are widely used in disaster detection and military reconnaissance and so on. However, their interpretation faces some challenges, e.g., deficiency of labeled data, inadequate utilization of data information and so on. In this paper, a complex-valued generative adversarial network (GAN) is proposed for the first time to address these issues. The complex number form of model complies with the physical mechanism of PolSAR data and in favor of utilizing and retaining amplitude and phase information of PolSAR data. GAN architecture and semi-supervised learning are combined to handle deficiency of labeled data. GAN expands training data and semi-supervised learning is used to train network with generated, labeled and unlabeled data. Experimental results on two benchmark data sets show that our model outperforms existing state-of-the-art models, especially for conditions with fewer labeled data.
We study the implicit bias of AdaGrad on separable linear classification problems. We show that AdaGrad converges to a direction that can be characterized as the solution of a quadratic optimization problem with the same feasible set as the hard SVM problem. We also give a discussion about how different choices of the hyperparameters of AdaGrad might impact this direction. This provides a deeper understanding of why adaptive methods do not seem to have the generalization ability as good as gradient descent does in practice.
Cane toads are invasive, toxic to native predators, compete with native insectivores, and have a devastating impact on Australian ecosystems, prompting the Australian government to list toads as a key threatening process under the Environment Protection and Biodiversity Conservation Act 1999. Mechanical cane toad traps could be made more native-fauna friendly if they could distinguish invasive cane toads from native species. Here we designed and trained a Convolution Neural Network (CNN) starting from the Xception CNN. The XToadGmp toad-recognition CNN we developed was trained end-to-end using heat-map Gaussian targets. After training, XToadGmp required minimum image pre/post-processing and when tested on 720x1280 shaped images, it achieved 97.1% classification accuracy on 1863 toad and 2892 not-toad test images, which were not used in training.