This paper presents a study of semi-supervised learning with large convolutional networks. We propose a pipeline, based on a teacher/student paradigm, that leverages a large collection of unlabelled images (up to 1 billion). Our main goal is to improve the performance for a given target architecture, like ResNet-50 or ResNext. We provide an extensive analysis of the success factors of our approach, which leads us to formulate some recommendations to produce high-accuracy models for image classification with semi-supervised learning. As a result, our approach brings important gains to standard architectures for image, video and fine-grained classification. For instance, by leveraging one billion unlabelled images, our learned vanilla ResNet-50 achieves 81.2% top-1 accuracy on the ImageNet benchmark.
The biological function of a protein stems from its 3-dimensional structure, which is thermodynamically determined by the energetics of interatomic forces between its amino acid building blocks (the order of amino acids, known as the sequence, defines a protein). Given the costs (time, money, human resources) of determining protein structures via experimental means such as X-ray crystallography, can we better describe and compare protein 3D structures in a robust and efficient manner, so as to gain meaningful biological insights? We begin by considering a relatively simple problem, limiting ourselves to just protein secondary structural elements. Historically, many computational methods have been devised to classify amino acid residues in a protein chain into one of several discrete secondary structures, of which the most well-characterized are the geometrically regular $\alpha$-helix and $\beta$-sheet; irregular structural patterns, such as 'turns' and 'loops', are less understood. Here, we present a study of Deep Learning techniques to classify the loop-like end cap structures which delimit $\alpha$-helices. Previous work used highly empirical and heuristic methods to manually classify helix capping motifs. Instead, we use structural data directly--including (i) backbone torsion angles computed from 3D structures, (ii) macromolecular feature sets (e.g., physicochemical properties), and (iii) helix cap classification data (from CAPS-DB)--as the ground truth to train a bidirectional long short-term memory (BiLSTM) model to classify helix cap residues. We tried different network architectures and scanned hyperparameters in order to train and assess several models; we also trained a Support Vector Classifier (SVC) to use as a baseline. Ultimately, we achieved 85% class-balanced accuracy with a deep BiLSTM model.
We empirically investigate the (negative) expected accuracy as an alternative loss function to cross entropy (negative log likelihood) for classification tasks. Coupled with softmax activation, it has small derivatives over most of its domain, and is therefore hard to optimize. A modified, leaky version is evaluated on a variety of classification tasks, including digit recognition, image classification, sequence tagging and tree tagging, using a variety of neural architectures such as logistic regression, multilayer perceptron, CNN, LSTM and Tree-LSTM. We show that it yields comparable or better accuracy compared to cross entropy. Furthermore, the proposed objective is shown to be more robust to label noise.
In this paper, we provide a comprehensive analysis of periocular-based sex-prediction (commonly referred to as gender classification) using state-of-the-art machine learning techniques. In order to reflect a more challenging scenario where periocular images are likely to be obtained from an unknown source, i.e. sensor, convolutional neural networks are trained on fused sets composed of several near-infrared (NIR) and visible wavelength (VW) image databases. In a cross-sensor scenario within each spectrum an average classification accuracy of approximately 85\% is achieved. When sex-prediction is performed across spectra an average classification accuracy of about 82\% is obtained. Finally, a multi-spectral sex-prediction yields a classification accuracy of 83\% on average. Compared to proposed works, obtained results provide a more realistic estimation of the feasibility to predict a subject's sex from the periocular region.
Soft biometric information such as gender can contribute to many applications like as identification and security. This paper explores the use of a Binary Statistical Features (BSIF) algorithm for classifying gender from iris texture images captured with NIR sensors. It uses the same pipeline for iris recognition systems consisting of iris segmentation, normalisation and then classification. Experiments show that applying BSIF is not straightforward since it can create artificial textures causing misclassification. In order to overcome this limitation, a new set of filters was trained from eye images and different sized filters with padding bands were tested on a subject-disjoint database. A Modified-BSIF (MBSIF) method was implemented. The latter achieved better gender classification results (94.6\% and 91.33\% for the left and right eye respectively). These results are competitive with the state of the art in gender classification. In an additional contribution, a novel gender labelled database was created and it will be available upon request.
The support vector machines (SVM) algorithm is a popular classification technique in data mining and machine learning. In this paper, we propose a distributed SVM algorithm and demonstrate its use in a number of applications. The algorithm is named high-performance support vector machines (HPSVM). The major contribution of HPSVM is two-fold. First, HPSVM provides a new way to distribute computations to the machines in the cloud without shuffling the data. Second, HPSVM minimizes the inter-machine communications in order to maximize the performance. We apply HPSVM to some real-world classification problems and compare it with the state-of-the-art SVM technique implemented in R on several public data sets. HPSVM achieves similar or better results.
Interpretable classifiers have recently witnessed an increase in attention from the data mining community because they are inherently easier to understand and explain than their more complex counterparts. Examples of interpretable classification models include decision trees, rule sets, and rule lists. Learning such models often involves optimizing hyperparameters, which typically requires substantial amounts of data and may result in relatively large models. In this paper, we consider the problem of learning compact yet accurate probabilistic rule lists for multiclass classification. Specifically, we propose a novel formalization based on probabilistic rule lists and the minimum description length (MDL) principle. This results in virtually parameter-free model selection that naturally allows to trade-off model complexity with goodness of fit, by which overfitting and the need for hyperparameter tuning are effectively avoided. Finally, we introduce the Classy algorithm, which greedily finds rule lists according to the proposed criterion. We empirically demonstrate that Classy selects small probabilistic rule lists that outperform state-of-the-art classifiers when it comes to the combination of predictive performance and interpretability. We show that Classy is insensitive to its only parameter, i.e., the candidate set, and that compression on the training set correlates with classification performance, validating our MDL-based selection criterion.
In this work we describe the preparation of a time series dataset of inertial measurements for determining the surface type under a wheeled robot. The data consists of over 7600 labeled time series samples, with the corresponding surface type annotation. This data was used in two public competitions with over 1500 participant in total. Additionally, we describe the performance of state-of-art deep learning models for time series classification, as well as propose a baseline model based on an ensemble of machine learning methods. The baseline achieves an accuracy of over 68% with our nine-category dataset.
Now that machine learning algorithms lie at the center of many resource allocation pipelines, computer scientists have been unwittingly cast as partial social planners. Given this state of affairs, important questions follow. What is the relationship between fairness as defined by computer scientists and notions of social welfare? In this paper, we present a welfare-based analysis of classification and fairness regimes. We translate a loss minimization program into a social welfare maximization problem with a set of implied welfare weights on individuals and groups--weights that can be analyzed from a distribution justice lens. In the converse direction, we ask what the space of possible labelings is for a given dataset and hypothesis class. We provide an algorithm that answers this question with respect to linear hyperplanes in $\mathbb{R}^d$ that runs in $O(n^dd)$. Our main findings on the relationship between fairness criteria and welfare center on sensitivity analyses of fairness-constrained empirical risk minimization programs. We characterize the ranges of $\Delta \epsilon$ perturbations to a fairness parameter $\epsilon$ that yield better, worse, and neutral outcomes in utility for individuals and by extension, groups. We show that applying more strict fairness criteria that are codified as parity constraints, can worsen welfare outcomes for both groups. More generally, always preferring "more fair" classifiers does not abide by the Pareto Principle---a fundamental axiom of social choice theory and welfare economics. Recent work in machine learning has rallied around these notions of fairness as critical to ensuring that algorithmic systems do not have disparate negative impact on disadvantaged social groups. By showing that these constraints often fail to translate into improved outcomes for these groups, we cast doubt on their effectiveness as a means to ensure justice.
Converting malware into images followed by vision-based deep learning algorithms has shown superior threat detection efficacy compared with classical machine learning algorithms. When malware are visualized as images, visual-based interpretation schemes can also be applied to extract insights of why individual samples are classified as malicious. In this work, via two case studies of dynamic malware classification, we extend the local interpretable model-agnostic explanation algorithm to explain image-based dynamic malware classification and examine its interpretation fidelity. For both case studies, we first train deep learning models via transfer learning on malware images, demonstrate high classification effectiveness, apply an explanation method on the images, and correlate the results back to the samples to validate whether the algorithmic insights are consistent with security domain expertise. In our first case study, the interpretation framework identifies indirect calls that uniquely characterize the underlying exploit behavior of a malware family. In our second case study, the interpretation framework extracts insightful information such as cryptography-related APIs when applied on images created from API existence, but generate ambiguous interpretation on images created from API sequences and frequencies. Our findings indicate that current image-based interpretation techniques are promising for explaining vision-based malware classification. We continue to develop image-based interpretation schemes specifically for security applications.
Competitive diving is a well recognized aquatic sport in which a person dives from a platform or a springboard into the water. Based on the acrobatics performed during the dive, diving is classified into a finite set of action classes which are standardized by FINA. In this work, we propose an attention guided LSTM-based neural network architecture for the task of diving classification. The network takes the frames of a diving video as input and determines its class. We evaluate the performance of the proposed model on a recently introduced competitive diving dataset, Diving48. It contains over 18000 video clips which covers 48 classes of diving. The proposed model outperforms the classification accuracy of the state-of-the-art models in both 2D and 3D frameworks by 11.54% and 4.24%, respectively. We show that the network is able to localize the diver in the video frames during the dive without being trained with such a supervision.
Deep learning has revolutionized computer vision utilizing the increased availability of big data and the power of parallel computational units such as graphical processing units. The vast majority of deep learning research is conducted using images as training data, however the biomedical domain is rich in physiological signals that are used for diagnosis and prediction problems. It is still an open research question how to best utilize signals to train deep neural networks. In this paper we define the term Signal2Image (S2Is) as trainable or non-trainable prefix modules that convert signals, such as Electroencephalography (EEG), to image-like representations making them suitable for training image-based deep neural networks defined as `base models'. We compare the accuracy and time performance of four S2Is (`signal as image', spectrogram, one and two layer Convolutional Neural Networks (CNNs)) combined with a set of `base models' (LeNet, AlexNet, VGGnet, ResNet, DenseNet) along with the depth-wise and 1D variations of the latter. We also provide empirical evidence that the one layer CNN S2I performs better in eleven out of fifteen tested models than non-trainable S2Is for classifying EEG signals and we present visual comparisons of the outputs of the S2Is.
We propose a sentiment classification method with a general machine learning framework. For feature representation, n-gram IDF is used to extract software-engineering-related, dataset-specific, positive, neutral, and negative n-gram expressions. For classifiers, an automated machine learning tool is used. In the comparison using publicly available datasets, our method achieved the highest F1 values in positive and negative sentences on all datasets.
We classify digits of real-world house numbers using convolutional neural networks (ConvNets). ConvNets are hierarchical feature learning neural networks whose structure is biologically inspired. Unlike many popular vision approaches that are hand-designed, ConvNets can automatically learn a unique set of features optimized for a given task. We augmented the traditional ConvNet architecture by learning multi-stage features and by using Lp pooling and establish a new state-of-the-art of 94.85% accuracy on the SVHN dataset (45.2% error improvement). Furthermore, we analyze the benefits of different pooling methods and multi-stage features in ConvNets. The source code and a tutorial are available at eblearn.sf.net.
In this paper we seek methods to effectively detect urban micro-events. Urban micro-events are events which occur in cities, have limited geographical coverage and typically affect only a small group of citizens. Because of their scale these are difficult to identify in most data sources. However, by using citizen sensing to gather data, detecting them becomes feasible. The data gathered by citizen sensing is often multimodal and, as a consequence, the information required to detect urban micro-events is distributed over multiple modalities. This makes it essential to have a classifier capable of combining them. In this paper we explore several methods of creating such a classifier, including early, late, hybrid fusion and representation learning using multimodal graphs. We evaluate performance on a real world dataset obtained from a live citizen reporting system. We show that a multimodal approach yields higher performance than unimodal alternatives. Furthermore, we demonstrate that our hybrid combination of early and late fusion with multimodal embeddings performs best in classification of urban micro-events.
Fairness has become a central issue for our research community as classification algorithms are adopted in societally critical domains such as recidivism prediction and loan approval. In this work, we consider the potential bias based on protected attributes (e.g., race and gender), and tackle this problem by learning latent representations of individuals that are statistically indistinguishable between protected groups while sufficiently preserving other information for classification. To do that, we develop a minimax adversarial framework with a generator to capture the data distribution and generate latent representations, and a critic to ensure that the distributions across different protected groups are similar. Our framework provides a theoretical guarantee with respect to statistical parity and individual fairness. Empirical results on four real-world datasets also show that the learned representation can effectively be used for classification tasks such as credit risk prediction while obstructing information related to protected groups, especially when removing protected attributes is not sufficient for fair classification.
Graph neural networks, which generalize deep neural network models to graph structured data, have attracted increasing attention in recent years. They usually learn node representations by transforming, propagating and aggregating node features and have been proven to improve the performance of many graph related tasks such as node classification and link prediction. To apply graph neural networks for the graph classification task, approaches to generate the \textit{graph representation} from node representations are demanded. A common way is to globally combine the node representations. However, rich structural information is overlooked. Thus a hierarchical pooling procedure is desired to preserve the graph structure during the graph representation learning. There are some recent works on hierarchically learning graph representation analogous to the pooling step in conventional convolutional neural (CNN) networks. However, the local structural information is still largely neglected during the pooling process. In this paper, we introduce a pooling operator $\pooling$ based on graph Fourier transform, which can utilize the node features and local structures during the pooling process. We then design pooling layers based on the pooling operator, which are further combined with traditional GCN convolutional layers to form a graph neural network framework $\m$ for graph classification. Theoretical analysis is provided to understand $\pooling$ from both local and global perspectives. Experimental results of the graph classification task on $6$ commonly used benchmarks demonstrate the effectiveness of the proposed framework.
Wide field small aperture telescopes are working horses for fast sky surveying. Transient discovery is one of their main tasks. Classification of candidate transient images between real sources and artifacts with high accuracy is an important step for transient discovery. In this paper, we propose two transient classification methods based on neural networks. The first method uses the convolutional neural network without pooling layers to classify transient images with low sampling rate. The second method assumes transient images as one dimensional signals and is based on recurrent neural networks with long short term memory and leaky ReLu activation function in each detection layer. Testing with real observation data, we find that although these two methods can both achieve more than 94% classification accuracy, they have different classification properties for different targets. Based on this result, we propose to use the ensemble learning method to further increase the classification accuracy to more than 97%.
Deep learning techniques are recently being used in fundus image analysis and diabetic retinopathy detection. Microaneurysms are an important indicator of diabetic retinopathy progression. We introduce a two-stage deep learning approach for microaneurysms segmentation using multiple scales of the input with selective sampling and embedding triplet loss. The model first segments on two scales and then the segmentations are refined with a classification model. To enhance the discriminative power of the classification model, we incorporate triplet embedding loss with a selective sampling routine. The model is evaluated quantitatively to assess the segmentation performance and qualitatively to analyze the model predictions. This approach introduces a 30.29% relative improvement over the fully convolutional neural network.
Vehicular Ad-hoc NETworks (VANET) can efficiently detect traffic congestion, but detection is not enough because congestion can be further classified as recurrent and non-recurrent congestion (NRC). In particular, NRC in an urban network is mainly caused by incidents, workzones, special events and adverse weather. We propose a framework for the real-time distributed classification of congestion into its components on a heterogeneous urban road network using VANET. We present models built on an understanding of the spatial and temporal causality measures and trained on synthetic data extended from a real case study of Cologne. Our performance evaluation shows a predictive accuracy of 87.63\% for the deterministic Classification Tree (CT), 88.83\% for the Naive Bayesian classifier (NB), 89.51\% for Random Forest (RF) and 89.17\% for the boosting technique. This framework can assist transportation agencies in reducing urban congestion by developing effective congestion mitigation strategies knowing the root causes of congestion.