We consider the problem of image classification for the purpose of aiding doctors in dermatological diagnosis. Dermatological diagnosis poses two major challenges for standard off-the-shelf techniques: First, the data distribution is typically extremely long tailed. Second, intra-class variability is often large. To address the first issue, we formulate the problem as low-shot learning, where once deployed, a base classifier must rapidly generalize to diagnose novel conditions given very few labeled examples. To model diverse classes effectively, we propose Prototypical Clustering Networks (PCN), an extension to Prototypical Networks that learns a mixture of prototypes for each class. Prototypes are initialized for each class via clustering and refined via an online update scheme. Classification is performed by measuring similarity to a weighted combination of prototypes within a class, where the weights are the inferred cluster responsibilities. We demonstrate the strengths of our approach in effective diagnosis on a realistic dataset of dermatological conditions.
Algorithmic discrimination is an important aspect when data is used for predictive purposes. This paper analyzes the relationships between discrimination and classification, data set partitioning, and decision models, as well as correlation. The paper uses real world data sets to demonstrate the existence of discrimination and the independence between the discrimination of data sets and the discrimination of classification models.
We investigate different strategies for automatic offensive language classification on German Twitter data. For this, we employ a sequentially combined BiLSTM-CNN neural network. Based on this model, three transfer learning tasks to improve the classification performance with background knowledge are tested. We compare 1. Supervised category transfer: social media data annotated with near-offensive language categories, 2. Weakly-supervised category transfer: tweets annotated with emojis they contain, 3. Unsupervised category transfer: tweets annotated with topic clusters obtained by Latent Dirichlet Allocation (LDA). Further, we investigate the effect of three different strategies to mitigate negative effects of 'catastrophic forgetting' during transfer learning. Our results indicate that transfer learning in general improves offensive language detection. Best results are achieved from pre-training our model on the unsupervised topic clustering of tweets in combination with thematic user cluster information.
In this paper, we propose an effective THresholding method based on ORder Statistic, called THORS, to convert an arbitrary scoring-type classifier, which can induce a continuous cumulative distribution function of the score, into a cost-sensitive one. The procedure, uses order statistic to find an optimal threshold for classification, requiring almost no knowledge of classifiers itself. Unlike common data-driven methods, we analytically show that THORS has theoretical guaranteed performance, theoretical bounds for the costs and lower time complexity. Coupled with empirical results on several real-world data sets, we argue that THORS is the preferred cost-sensitive technique.
We examine two fundamental tasks associated with graph representation learning: link prediction and node classification. We present a new autoencoder architecture capable of learning a joint representation of local graph structure and available node features for the simultaneous multi-task learning of unsupervised link prediction and semi-supervised node classification. Our simple, yet effective and versatile model is efficiently trained end-to-end in a single stage, whereas previous related deep graph embedding methods require multiple training steps that are difficult to optimize. We provide an empirical evaluation of our model on five benchmark relational, graph-structured datasets and demonstrate significant improvement over three strong baselines for graph representation learning. Reference code and data are available at https://github.com/vuptran/graph-representation-learning
With the rapid development of deep learning, there have been an unprecedentedly large number of trained deep network models available online. Reusing such trained models can significantly reduce the cost of training the new models from scratch, if not infeasible at all as the annotations used for the training original networks are often unavailable to public. We propose in this paper to study a new model-reusing task, which we term as \emph{knowledge amalgamation}. Given multiple trained teacher networks, each of which specializes in a different classification problem, the goal of knowledge amalgamation is to learn a lightweight student model capable of handling the comprehensive classification. We assume no other annotations except the outputs from the teacher models are available, and thus focus on extracting and amalgamating knowledge from the multiple teachers. To this end, we propose a pilot two-step strategy to tackle the knowledge amalgamation task, by learning first the compact feature representations from teachers and then the network parameters in a layer-wise manner so as to build the student model. We apply this approach to four public datasets and obtain very encouraging results: even without any human annotation, the obtained student model is competent to handle the comprehensive classification task and in most cases outperforms the teachers in individual sub-tasks.
We propose and study a new projection formula for training binary weight convolutional neural networks. The projection formula measures the error in approximating a full precision (32 bit) vector by a 1-bit vector in the l_1 norm instead of the standard l_2 norm. The l_1 projector is in closed analytical form and involves a median computation instead of an arithmatic average in the l_2 projector. Experiments on 10 keywords classification show that the l_1 (median) BinaryConnect (BC) method outperforms the regular BC, regardless of cold or warm start. The binary network trained by median BC and a recent blending technique reaches test accuracy 92.4%, which is 1.1% lower than the full-precision network accuracy 93.5%. On Android phone app, the trained binary network doubles the speed of full-precision network in spoken keywords recognition.
In this paper, we introduce Deep Probabilistic Ensembles (DPEs), a scalable technique that uses a regularized ensemble to approximate a deep Bayesian Neural Network (BNN). We do so by incorporating a KL divergence penalty term into the training objective of an ensemble, derived from the evidence lower bound used in variational inference. We evaluate the uncertainty estimates obtained from our models for active learning on visual classification. Our approach steadily improves upon active learning baselines as the annotation budget is increased.
Recent advances in deep learning have achieved impressive gains in classification accuracy on a variety of types of data, including images and text. Despite these gains, however, concerns have been raised about the calibration, robustness, and interpretability of these models. In this paper we propose a simple way to modify any conventional deep architecture to automatically provide more transparent explanations for classification decisions, as well as an intuitive notion of the credibility of each prediction. Specifically, we draw on ideas from nonparametric kernel regression, and propose to predict labels based on a weighted sum of training instances, where the weights are determined by distance in a learned instance-embedding space. Working within the framework of conformal methods, we propose a new measure of nonconformity suggested by our model, and experimentally validate the accompanying theoretical expectations, demonstrating improved transparency, controlled error rates, and robustness to out-of-domain data, without compromising on accuracy or calibration.
Extreme multi-label text classification (XMTC) is an important problem in the era of big data, for tagging a given text with the most relevant multiple labels from an extremely large-scale label set. XMTC can be found in many applications, such as item categorization, web page tagging, and news annotation. Traditionally most methods used bag-of-words (BOW) as inputs, ignoring word context as well as deep semantic information. Recent attempts to overcome the problems of BOW by deep learning still suffer from 1) failing to capture the important subtext for each label and 2) lack of scalability against the huge number of labels. We propose a new label tree-based deep learning model for XMTC, called AttentionXML, with two unique features: 1) a multi-label attention mechanism with raw text as input, which allows to capture the most relevant part of text to each label; and 2) a shallow and wide probabilistic label tree (PLT), which allows to handle millions of labels, especially for "tail labels". We empirically compared the performance of AttentionXML with those of eight state-of-the-art methods over six benchmark datasets, including Amazon-3M with around 3 million labels. AttentionXML outperformed all competing methods under all experimental settings. Experimental results also show that AttentionXML achieved the best performance against tail labels among label tree-based methods. The code and datasets are available at http://github.com/yourh/AttentionXML .
tl;dr: no, it cannot, at least not on average on the standard archive problems. We assess whether using six smoothing algorithms (moving average, exponential smoothing, Gaussian filter, Savitzky-Golay filter, Fourier approximation and a recursive median sieve) could be automatically applied to time series classification problems as a preprocessing step to improve the performance of three benchmark classifiers (1-Nearest Neighbour with Euclidean and Dynamic Time Warping distances, and Rotation Forest). We found no significant improvement over unsmoothed data even when we set the smoothing parameter through cross validation. We are not claiming smoothing has no worth. It has an important role in exploratory analysis and helps with specific classification problems where domain knowledge can be exploited. What we observe is that the automatic application does not help and that we cannot explain the improvement of other time series classification algorithms over the baseline classifiers simply as a function of the absence of smoothing.
The ability to correctly classify and retrieve apparel images has a variety of applications important to e-commerce, online advertising and internet search. In this work, we propose a robust framework for fine-grained apparel classification, in-shop and cross-domain retrieval which eliminates the requirement of rich annotations like bounding boxes and human-joints or clothing landmarks, and training of bounding box/ key-landmark detector for the same. Factors such as subtle appearance differences, variations in human poses, different shooting angles, apparel deformations, and self-occlusion add to the challenges in classification and retrieval of apparel items. Cross-domain retrieval is even harder due to the presence of large variation between online shopping images, usually taken in ideal lighting, pose, positive angle and clean background as compared with street photos captured by users in complicated conditions with poor lighting and cluttered scenes. Our framework uses compact bilinear CNN with tensor sketch algorithm to generate embeddings that capture local pairwise feature interactions in a translationally invariant manner. For apparel classification, we pass the feature embeddings through a softmax classifier, while, the in-shop and cross-domain retrieval pipelines use a triplet-loss based optimization approach, such that squared Euclidean distance between embeddings measures the dissimilarity between the images. Unlike previous works that relied on bounding box, key clothing landmarks or human joint detectors to assist the final deep classifier, proposed framework can be trained directly on the provided category labels or generated triplets for triplet loss optimization. Lastly, Experimental results on the DeepFashion fine-grained categorization, and in-shop and consumer-to-shop retrieval datasets provide a comparative analysis with previous work performed in the domain.
In biomedical research, many different types of patient data can be collected, such as various types of omics data and medical imaging modalities. Applying multi-view learning to these different sources of information can increase the accuracy of medical classification models compared with single-view procedures. However, collecting biomedical data can be expensive and/or burdening for patients, so that it is important to reduce the amount of required data collection. It is therefore necessary to develop multi-view learning methods which can accurately identify those views that are most important for prediction. In recent years, several biomedical studies have used an approach known as multi-view stacking (MVS), where a model is trained on each view separately and the resulting predictions are combined through stacking. In these studies, MVS has been shown to increase classification accuracy. However, the MVS framework can also be used for selecting a subset of important views. To study the view selection potential of MVS, we develop a special case called stacked penalized logistic regression (StaPLR). Compared with existing view-selection methods, StaPLR can make use of faster optimization algorithms and is easily parallelized. We show that nonnegativity constraints on the parameters of the function which combines the views play an important role in preventing unimportant views from entering the model. We investigate the performance of StaPLR through simulations, and consider two real data examples. We compare the performance of StaPLR with an existing view selection method called the group lasso and observe that, in terms of view selection, StaPLR is often more conservative and has a consistently lower false positive rate.
Classification, the process of assigning a label (or class) to an observation given its features, is a common task in many applications. Nonetheless in most real-life applications, the labels can not be fully explained by the observed features. Indeed there can be many factors hidden to the modellers. The unexplained variation is then treated as some random noise which is handled differently depending on the method retained by the practitioner. This work focuses on two simple and widely used supervised classification algorithms: discrete choice models and artificial neural networks in the context of binary classification. Through various numerical experiments involving continuous or discrete explanatory features, we present a comparison of the retained methods' performance in presence of missing variables. The impact of the distribution of the two classes in the training data is also investigated. The outcomes of those experiments highlight the fact that artificial neural networks outperforms the discrete choice models, except when the distribution of the classes in the training data is highly unbalanced. Finally, this work provides some guidelines for choosing the right classifier with respect to the training data.
This paper introduces a novel method for the representation of images that is semantic by nature, addressing the question of computation intelligibility in computer vision tasks. More specifically, our proposition is to introduce what we call a semantic bottleneck in the processing pipeline, which is a crossing point in which the representation of the image is entirely expressed with natural language , while retaining the efficiency of numerical representations. We show that our approach is able to generate semantic representations that give state-of-the-art results on semantic content-based image retrieval and also perform very well on image classification tasks. Intelligibility is evaluated through user centered experiments for failure detection.
This paper describes the CIS slot filling system for the TAC Cold Start evaluations 2015. It extends and improves the system we have built for the evaluation last year. This paper mainly describes the changes to our last year's system. Especially, it focuses on the coreference and classification component. For coreference, we have performed several analysis and prepared a resource to simplify our end-to-end system and improve its runtime. For classification, we propose to use neural networks. We have trained convolutional and recurrent neural networks and combined them with traditional evaluation methods, namely patterns and support vector machines. Our runs for the 2015 evaluation have been designed to directly assess the effect of each network on the end-to-end performance of the system. The CIS system achieved rank 3 of all slot filling systems participating in the task.
This paper introduces the 3DCapsule, which is a 3D extension of the recently introduced Capsule concept that makes it applicable to unordered point sets. The original Capsule relies on the existence of a spatial relationship between the elements in the feature map it is presented with, whereas in point permutation invariant formulations of 3D point set classification methods, such relationships are typically lost. Here, a new layer called ComposeCaps is introduced that, in lieu of a spatially relevant feature mapping, learns a new mapping that can be exploited by the 3DCapsule. Previous works in the 3D point set classification domain have focused on other parts of the architecture, whereas instead, the 3DCapsule is a drop-in replacement of the commonly used fully connected classifier. It is demonstrated via an ablation study, that when the 3DCapsule is applied to recent 3D point set classification architectures, it consistently shows an improvement, in particular when subjected to noisy data. Similarly, the ComposeCaps layer is evaluated and demonstrates an improvement over the baseline. In an apples-to-apples comparison against state-of-the-art methods, again, better performance is demonstrated by the 3DCapsule.
We propose a recurrent neural network classifier to detect pathologies in 12-lead ECG signals and train and validate the classifier with the Chinese physiological signal challenge dataset (http://www.icbeb.org/Challenge.html). The recurrent neural network consists of two bi-directional LSTM layers and can train on arbitrary-length ECG signals. Our best trained model achieved an average F1 score of 74.15% on the validation set. Keywords: ECG classification, Deep learning, RNN, Bi-directional LSTM, QRS detection.
We present a general theory of Group equivariant Convolutional Neural Networks (G-CNNs) on homogeneous spaces such as Euclidean space and the sphere. Feature maps in these networks represent fields on a homogeneous base space, and layers are equivariant maps between spaces of fields. The theory enables a systematic classification of all existing G-CNNs in terms of their symmetry group, base space, and field type. We also consider a fundamental question: what is the most general kind of equivariant linear map between feature spaces (fields) of given types? Following Mackey, we show that such maps correspond one-to-one with convolutions using equivariant kernels, and characterize the space of such kernels.
The abundance of text data being produced in the modern age makes it increasingly important to intuitively group, categorize, or classify text data by theme for efficient retrieval and search. Yet, the high dimensionality and imprecision of text data, or more generally language as a whole, prove to be challenging when attempting to perform unsupervised document clustering. In this thesis, we present two novel methods for improving unsupervised document clustering/classification by theme. The first is to improve document representations. We look to exploit "term neighborhoods" and "blur" semantic weight across neighboring terms. These neighborhoods are located in the semantic space afforded by "word embeddings." The second method is for cluster revision, based on what we deem as "stochastic barcoding", or "S- Barcode" patterns. Text data is inherently high dimensional, yet clustering typically takes place in a low dimensional representation space. Our method utilizes lower dimension clustering results as initial cluster configurations, and iteratively revises the configuration in the high dimensional space. We show with experimental results how both of the two methods improve the quality of document clustering. While this thesis elaborates on the two new conceptual contributions, a joint thesis by David Yan details the feature transformation and software architecture we developed for unsupervised document classification.