Image blur and image noise are common distortions during image acquisition. In this paper, we systematically study the effect of image distortions on the deep neural network (DNN) image classifiers. First, we examine the DNN classifier performance under four types of distortions. Second, we propose two approaches to alleviate the effect of image distortion: re-training and fine-tuning with noisy images. Our results suggest that, under certain conditions, fine-tuning with noisy images can alleviate much effect due to distorted inputs, and is more practical than re-training.
We introduce SalGAN, a deep convolutional neural network for visual saliency prediction trained with adversarial examples. The first stage of the network consists of a generator model whose weights are learned by back-propagation computed from a binary cross entropy (BCE) loss over downsampled versions of the saliency maps. The resulting prediction is processed by a discriminator network trained to solve a binary classification task between the saliency maps generated by the generative stage and the ground truth ones. Our experiments show how adversarial training allows reaching state-of-the-art performance across different metrics when combined with a widely-used loss function like BCE. Our results can be reproduced with the source code and trained models available at https://imatge-upc.github.io/saliency-salgan-2017/.
Classifying the various shapes and attributes of a glioma cell nucleus is crucial for diagnosis and understanding the disease. We investigate automated classification of glioma nuclear shapes and visual attributes using Convolutional Neural Networks (CNNs) on pathology images of automatically segmented nuclei. We propose three methods that improve the performance of a previously-developed semi-supervised CNN. First, we propose a method that allows the CNN to focus on the most important part of an image- the image's center containing the nucleus. Second, we inject (concatenate) pre-extracted VGG features into an intermediate layer of our Semi-Supervised CNN so that during training, the CNN can learn a set of complementary features. Third, we separate the losses of the two groups of target classes (nuclear shapes and attributes) into a single-label loss and a multi-label loss so that the prior knowledge of inter-label exclusiveness can be incorporated. On a dataset of 2078 images, the proposed methods combined reduce the error rate of attribute and shape classification by 21.54% and 15.07% respectively compared to the existing state-of-the-art method on the same dataset.
In this paper, we explore ordinal classification (in the context of deep neural networks) through a simple modification of the squared error loss which not only allows it to not only be sensitive to class ordering, but also allows the possibility of having a discrete probability distribution over the classes. Our formulation is based on the use of a softmax hidden layer, which has received relatively little attention in the literature. We empirically evaluate its performance on the Kaggle diabetic retinopathy dataset, an ordinal and high-resolution dataset and show that it outperforms all of the baselines employed.
Convolutional Neural Networks (CNNs) have proven very effective in image classification and show promise for audio. We use various CNN architectures to classify the soundtracks of a dataset of 70M training videos (5.24 million hours) with 30,871 video-level labels. We examine fully connected Deep Neural Networks (DNNs), AlexNet [1], VGG [2], Inception [3], and ResNet [4]. We investigate varying the size of both training set and label vocabulary, finding that analogs of the CNNs used in image classification do well on our audio classification task, and larger training and label sets help up to a point. A model using embeddings from these classifiers does much better than raw features on the Audio Set [5] Acoustic Event Detection (AED) classification task.
We propose an algorithm to uncover the intrinsic low-rank component of a high-dimensional, graph-smooth and grossly-corrupted dataset, under the situations that the underlying graph is unknown. Based on a model with a low-rank component plus a sparse perturbation, and an initial graph estimation, our proposed algorithm simultaneously learns the low-rank component and refines the graph. Our evaluations using synthetic and real brain imaging data in unsupervised and supervised classification tasks demonstrate encouraging performance.
Crowdsourcing systems are popular for solving large-scale labelling tasks with low-paid workers. We study the problem of recovering the true labels from the possibly erroneous crowdsourced labels under the popular Dawid-Skene model. To address this inference problem, several algorithms have recently been proposed, but the best known guarantee is still significantly larger than the fundamental limit. We close this gap by introducing a tighter lower bound on the fundamental limit and proving that Belief Propagation (BP) exactly matches this lower bound. The guaranteed optimality of BP is the strongest in the sense that it is information-theoretically impossible for any other algorithm to correctly label a larger fraction of the tasks. Experimental results suggest that BP is close to optimal for all regimes considered and improves upon competing state-of-the-art algorithms.
Deriving from the gradient vector of a generative model of local features, Fisher vector coding (FVC) has been identified as an effective coding method for image classification. Most, if not all, FVC implementations employ the Gaussian mixture model (GMM) to depict the generation process of local features. However, the representative power of the GMM could be limited because it essentially assumes that local features can be characterized by a fixed number of feature prototypes and the number of prototypes is usually small in FVC. To handle this limitation, in this paper we break the convention which assumes that a local feature is drawn from one of few Gaussian distributions. Instead, we adopt a compositional mechanism which assumes that a local feature is drawn from a Gaussian distribution whose mean vector is composed as the linear combination of multiple key components and the combination weight is a latent random variable. In this way, we can greatly enhance the representative power of the generative model of FVC. To implement our idea, we designed two particular generative models with such a compositional mechanism.
The classification of MRI images according to the anatomical field of view is a necessary task to solve when faced with the increasing quantity of medical images. In parallel, advances in deep learning makes it a suitable tool for computer vision problems. Using a common architecture (such as AlexNet) provides quite good results, but not sufficient for clinical use. Improving the model is not an easy task, due to the large number of hyper-parameters governing both the architecture and the training of the network, and to the limited understanding of their relevance. Since an exhaustive search is not tractable, we propose to optimize the network first by random search, and then by an adaptive search based on Gaussian Processes and Probability of Improvement. Applying this method on a large and varied MRI dataset, we show a substantial improvement between the baseline network and the final one (up to 20\% for the most difficult classes).
In this paper we propose a new document classification method, bridging discrepancies (so-called semantic gap) between the training set and the application sets of textual data. We demonstrate its superiority over classical text classification approaches, including traditional classifier ensembles. The method consists in combining a document categorization technique with a single classifier or a classifier ensemble (SEMCOM algorithm - Committee with Semantic Categorizer).
We introduce a library of geometric voxel features for CAD surface recognition/retrieval tasks. Our features include local versions of the intrinsic volumes (the usual 3D volume, surface area, integrated mean and Gaussian curvature) and a few closely related quantities. We also compute Haar wavelet and statistical distribution features by aggregating raw voxel features. We apply our features to object classification on the ESB data set and demonstrate accurate results with a small number of shallow decision trees.
We consider the task of identifying attitudes towards a given set of entities from text. Conventionally, this task is decomposed into two separate subtasks: target detection that identifies whether each entity is mentioned in the text, either explicitly or implicitly, and polarity classification that classifies the exact sentiment towards an identified entity (the target) into positive, negative, or neutral. Instead, we show that attitude identification can be solved with an end-to-end machine learning architecture, in which the two subtasks are interleaved by a deep memory network. In this way, signals produced in target detection provide clues for polarity classification, and reversely, the predicted polarity provides feedback to the identification of targets. Moreover, the treatments for the set of targets also influence each other -- the learned representations may share the same semantics for some targets but vary for others. The proposed deep memory network, the AttNet, outperforms methods that do not consider the interactions between the subtasks or those among the targets, including conventional machine learning methods and the state-of-the-art deep learning models.
This work presents a fast and scalable algorithm for incremental learning of Gaussian mixture models. By performing rank-one updates on its precision matrices and determinants, its asymptotic time complexity is of \BigO{NKD^2} for $N$ data points, $K$ Gaussian components and $D$ dimensions. The resulting algorithm can be applied to high dimensional tasks, and this is confirmed by applying it to the classification datasets MNIST and CIFAR-10. Additionally, in order to show the algorithm's applicability to function approximation and control tasks, it is applied to three reinforcement learning tasks and its data-efficiency is evaluated.
Recent successes in learning-based image classification, however, heavily rely on the large number of annotated training samples, which may require considerable human efforts. In this paper, we propose a novel active learning framework, which is capable of building a competitive classifier with optimal feature representation via a limited amount of labeled training instances in an incremental learning manner. Our approach advances the existing active learning methods in two aspects. First, we incorporate deep convolutional neural networks into active learning. Through the properly designed framework, the feature representation and the classifier can be simultaneously updated with progressively annotated informative samples. Second, we present a cost-effective sample selection strategy to improve the classification performance with less manual annotations. Unlike traditional methods focusing on only the uncertain samples of low prediction confidence, we especially discover the large amount of high confidence samples from the unlabeled set for feature learning. Specifically, these high confidence samples are automatically selected and iteratively assigned pseudo-labels. We thus call our framework "Cost-Effective Active Learning" (CEAL) standing for the two advantages.Extensive experiments demonstrate that the proposed CEAL framework can achieve promising results on two challenging image classification datasets, i.e., face recognition on CACD database [1] and object categorization on Caltech-256 [2].
Learning to navigate in complex environments with dynamic elements is an important milestone in developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency and task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. In particular we consider jointly learning the goal-driven reinforcement learning problem with auxiliary depth prediction and loop closure classification tasks. This approach can learn to navigate from raw sensory input in complicated 3D mazes, approaching human-level performance even under conditions where the goal location changes frequently. We provide detailed analysis of the agent behaviour, its ability to localise, and its network activity dynamics, showing that the agent implicitly learns key navigation abilities.
Inverse classification is the process of perturbing an instance in a meaningful way such that it is more likely to conform to a specific class. Historical methods that address such a problem are often framed to leverage only a single classifier, or specific set of classifiers. These works are often accompanied by naive assumptions. In this work we propose generalized inverse classification (GIC), which avoids restricting the classification model that can be used. We incorporate this formulation into a refined framework in which GIC takes place. Under this framework, GIC operates on features that are immediately actionable. Each change incurs an individual cost, either linear or non-linear. Such changes are subjected to occur within a specified level of cumulative change (budget). Furthermore, our framework incorporates the estimation of features that change as a consequence of direct actions taken (indirectly changeable features). To solve such a problem, we propose three real-valued heuristic-based methods and two sensitivity analysis-based comparison methods, each of which is evaluated on two freely available real-world datasets. Our results demonstrate the validity and benefits of our formulation, framework, and methods.
Most high-dimensional estimation and prediction methods propose to minimize a cost function (empirical risk) that is written as a sum of losses associated to each data point. In this paper we focus on the case of non-convex losses, which is practically important but still poorly understood. Classical empirical process theory implies uniform convergence of the empirical risk to the population risk. While uniform convergence implies consistency of the resulting M-estimator, it does not ensure that the latter can be computed efficiently. In order to capture the complexity of computing M-estimators, we propose to study the landscape of the empirical risk, namely its stationary points and their properties. We establish uniform convergence of the gradient and Hessian of the empirical risk to their population counterparts, as soon as the number of samples becomes larger than the number of unknown parameters (modulo logarithmic factors). Consequently, good properties of the population risk can be carried to the empirical risk, and we can establish one-to-one correspondence of their stationary points. We demonstrate that in several problems such as non-convex binary classification, robust regression, and Gaussian mixture model, this result implies a complete characterization of the landscape of the empirical risk, and of the convergence properties of descent algorithms. We extend our analysis to the very high-dimensional setting in which the number of parameters exceeds the number of samples, and provide a characterization of the empirical risk landscape under a nearly information-theoretically minimal condition. Namely, if the number of samples exceeds the sparsity of the unknown parameters vector (modulo logarithmic factors), then a suitable uniform convergence result takes place. We apply this result to non-convex binary classification and robust regression in very high-dimension.
Sentiments expressed in user-generated short text and sentences are nuanced by subtleties at lexical, syntactic, semantic and pragmatic levels. To address this, we propose to augment traditional features used for sentiment analysis and sarcasm detection, with cognitive features derived from the eye-movement patterns of readers. Statistical classification using our enhanced feature set improves the performance (F-score) of polarity detection by a maximum of 3.7% and 9.3% on two datasets, over the systems that use only traditional features. We perform feature significance analysis, and experiment on a held-out dataset, showing that cognitive features indeed empower sentiment analyzers to handle complex constructs.
For argumentation mining, there are several sub-tasks such as argumentation component type classification, relation classification. Existing research tends to solve such sub-tasks separately, but ignore the close relation between them. In this paper, we present a joint framework incorporating logical relation between sub-tasks to improve the performance of argumentation structure generation. We design an objective function to combine the predictions from individual models for each sub-task and solve the problem with some constraints constructed from background knowledge. We evaluate our proposed model on two public corpora and the experiment results show that our model can outperform the baseline that uses a separate model significantly for each sub-task. Our model also shows advantages on component-related sub-tasks compared to a state-of-the-art joint model based on the evidence graph.
In this paper we discuss the stability properties of convolutional neural networks. Convolutional neural networks are widely used in machine learning. In classification they are mainly used as feature extractors. Ideally, we expect similar features when the inputs are from the same class. That is, we hope to see a small change in the feature vector with respect to a deformation on the input signal. This can be established mathematically, and the key step is to derive the Lipschitz properties. Further, we establish that the stability results can be extended for more general networks. We give a formula for computing the Lipschitz bound, and compare it with other methods to show it is closer to the optimal value.