PyTorchKR Papers

최신 논문

총 11,809편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1911.03011

Adaptive Kernel Value Caching for SVM Training

Support Vector Machines (SVMs) can solve structured multi-output learning problems such as multi-label classification, multiclass classification and vector regression. SVM training is expensive especially for large and high dimensional datasets. The bottleneck of the SVM training often lies in the kernel value computation. In many real-world problems, the same kernel values are used in many iterations during the training, which makes the caching of kernel values potentially useful. The majority of the existing studies simply adopt the LRU (least recently used) replacement strategy for caching kernel values. However, as we analyze in this paper, the LRU strategy generally achieves high hit ratio near the final stage of the training, but does not work well in the whole training process. Therefore, we propose a new caching strategy called EFU (less frequently used) which replaces the less frequently used kernel values that enhances LFU (least frequently used). Our experimental results show that EFU often has 20\% higher hit ratio than LRU in the training with the Gaussian kernel. To further optimize the strategy, we propose a caching strategy called HCST (hybrid caching for the SVM training), which has a novel mechanism to automatically adapt the better caching strategy in the different stages of the training. We have integrated the caching strategy into ThunderSVM, a recent SVM library on many-core processors. Our experiments show that HCST adaptively achieves high hit ratios with little runtime overhead among different problems including multi-label classification, multiclass classification and regression problems. Compared with other existing caching strategies, HCST achieves 20\% more reduction in training time on average.

논문 ID: document/7298998

Fisher Vectors Meet Neural Networks: A Hybrid Classification Architecture

Fisher Vectors (FV) and Convolutional Neural Networks(CNN) are two image classification pipelines with different strengths. While CNNs have shown superior accuracy on a number of classification tasks, FV classifiers are typically less costly to train and evaluate. We propose a hybrid architecture that combines their strengths: the first unsupervised layers rely on the FV while the subsequent fully-connected supervised layers are trained with back-propagation. We show experimentally that this hybrid architecture significantly outperforms standard FV systems without incurring the high cost that comes with CNNs. We also derive competitive mid-level features from our architecture that are readily applicable to other class sets and even to new tasks.

arXiv:1911.02237

Localization-aware Channel Pruning for Object Detection

Channel pruning is one of the important methods for deep model compression. Most of existing pruning methods mainly focus on classification. Few of them conduct systematic research on object detection. However, object detection is different from classification, which requires not only semantic information but also localization information. In this paper, based on discrimination-aware channel pruning (DCP) which is state-of-the-art pruning method for classification, we propose a localization-aware auxiliary network to find out the channels with key information for classification and regression so that we can conduct channel pruning directly for object detection, which saves lots of time and computing resources. In order to capture the localization information, we first design the auxiliary network with a contextual ROIAlign layer which can obtain precise localization information of the default boxes by pixel alignment and enlarges the receptive fields of the default boxes when pruning shallow layers. Then, we construct a loss function for object detection task which tends to keep the channels that contain the key information for classification and regression. Extensive experiments demonstrate the effectiveness of our method. On MS COCO, we prune 70\% parameters of the SSD based on ResNet-50 with modest accuracy drop, which outperforms the-state-of-art method.

arXiv:1911.02098

A Scalable Multilabel Classification to Deploy Deep Learning Architectures For Edge Devices

Convolution Neural Networks (CNN) have performed well in many applications such as object detection, pattern recognition, video surveillance and so on. CNN carryout feature extraction on labelled data to perform classification. Multi-label classification assigns more than one label to a particular data sample in a data set. In multi-label classification, properties of a data point that are considered to be mutually exclusive are classified. However, existing multi-label classification requires some form of data pre-processing that involves image training data cropping or image tiling. The computation and memory requirement of these multi-label CNN models makes their deployment on edge devices challenging. In this paper, we propose a methodology that solves this problem by extending the capability of existing multi-label classification and provide models with lower latency that requires smaller memory size when deployed on edge devices. We make use of a single CNN model designed with multiple loss layers and multiple accuracy layers. This methodology is tested on state-of-the-art deep learning algorithms such as AlexNet, GoogleNet and SqueezeNet using the Stanford Cars Dataset and deployed on Raspberry Pi3. From the results the proposed methodology achieves comparable accuracy with 1.8x less MACC operation, 0.97x reduction in latency and 0.5x, 0.84x and 0.97x reduction in size for the generated AlexNet, GoogleNet and SqueezeNet CNN models respectively when compared to conventional ways of achieving multi-label classification like hard-coding multi-label instances into single labels. The methodology also yields CNN models that achieve 50\% less MACC operations, 50% reduction in latency and size of generated versions of AlexNet, GoogleNet and SqueezeNet respectively when compared to conventional ways using 2 different single-labelled models to achieve multi-label classification.

arXiv:1911.01468

Auditing and Achieving Intersectional Fairness in Classification Problems

Machine learning algorithms are extensively used to make increasingly more consequential decisions about people, so achieving optimal predictive performance can no longer be the only focus. A particularly important consideration is fairness with respect to race, gender, or any other sensitive attribute. This paper studies intersectional fairness, where intersections of multiple sensitive attributes are considered. Prior research has mainly focused on fairness with respect to a single sensitive attribute, with intersectional fairness being comparatively less studied despite its critical importance for the safety of modern machine learning systems. We present a comprehensive framework for auditing and achieving intersectional fairness in classification problems: we define a suite of metrics to assess intersectional fairness in the data or model outputs by extending known single-attribute fairness metrics, and propose methods for robustly estimating them even when some intersectional subgroups are underrepresented. Furthermore, we develop post-processing techniques to mitigate any detected intersectional bias in a classification model. Our techniques do not rely on any assumptions regarding the underlying model and preserve predictive performance at a guaranteed level of fairness. Finally, we give guidance on a practical implementation, showing how the proposed methods perform on a real-world dataset.

arXiv:1911.01898

3D Deformable Convolutions for MRI classification

Deep learning convolutional neural networks have proved to be a powerful tool for MRI analysis. In current work, we explore the potential of the deformable convolutional deep neural network layers for MRI data classification. We propose new 3D deformable convolutions(d-convolutions), implement them in VoxResNet architecture and apply for structural MRI data classification. We show that 3D d-convolutions outperform standard ones and are effective for unprocessed 3D MR images being robust to particular geometrical properties of the data. Firstly proposed dVoxResNet architecture exhibits high potential for the use in MRI data classification.

arXiv:1911.01026공식 코드 · Papers with Code

Metric Learning for Dynamic Text Classification

Traditional text classifiers are limited to predicting over a fixed set of labels. However, in many real-world applications the label set is frequently changing. For example, in intent classification, new intents may be added over time while others are removed. We propose to address the problem of dynamic text classification by replacing the traditional, fixed-size output layer with a learned, semantically meaningful metric space. Here the distances between textual inputs are optimized to perform nearest-neighbor classification across overlapping label sets. Changing the label set does not involve removing parameters, but rather simply adding or removing support points in the metric space. Then the learned metric can be fine-tuned with only a few additional training examples. We demonstrate that this simple strategy is robust to changes in the label space. Furthermore, our results show that learning a non-Euclidean metric can improve performance in the low data regime, suggesting that further work on metric spaces may benefit low-resource research.

arXiv:1911.00896

Generalized Learning with Rejection for Classification and Regression Problems

Learning with rejection (LWR) allows development of machine learning systems with the ability to discard low confidence decisions generated by a prediction model. That is, just like human experts, LWR allows machine models to abstain from generating a prediction when reliability of the prediction is expected to be low. Several frameworks for this learning with rejection have been proposed in the literature. However, most of them work for classification problems only and regression with rejection has not been studied in much detail. In this work, we present a neural framework for LWR based on a generalized meta-loss function that involves simultaneous training of two neural network models: a predictor model for generating predictions and a rejecter model for deciding whether the prediction should be accepted or rejected. The proposed framework can be used for classification as well as regression and other related machine learning tasks. We have demonstrated the applicability and effectiveness of the method on synthetically generated data as well as benchmark datasets from UCI machine learning repository for both classification and regression problems. Despite being simpler in implementation, the proposed scheme for learning with rejection has shown to perform at par or better than previously proposed methods. Furthermore, we have applied the method to the problem of hurricane intensity prediction from satellite imagery. Significant improvement in performance as compared to conventional supervised methods shows the effectiveness of the proposed scheme in real-world regression problems.

arXiv:1911.01071

Supervised level-wise pretraining for recurrent neural network initialization in multi-class classification

Recurrent Neural Networks (RNNs) can be seriously impacted by the initial parameters assignment, which may result in poor generalization performances on new unseen data. With the objective to tackle this crucial issue, in the context of RNN based classification, we propose a new supervised layer-wise pretraining strategy to initialize network parameters. The proposed approach leverages a data-aware strategy that sets up a taxonomy of classification problems automatically derived by the model behavior. To the best of our knowledge, despite the great interest in RNN-based classification, this is the first data-aware strategy dealing with the initialization of such models. The proposed strategy has been tested on four benchmarks coming from two different domains, i.e., Speech Recognition and Remote Sensing. Results underline the significance of our approach and point out that data-aware strategies positively support the initialization of Recurrent Neural Network based classification models.

arXiv:1911.00155공식 코드 · Papers with Code

Centroid Based Concept Learning for RGB-D Indoor Scene Classification

This paper contributes a novel cognitively-inspired method for RGB-D indoor scene classification. High intra-class variance and low inter-class variance make indoor scene classification an extremely challenging task. To cope with this problem, we propose a clustering approach inspired by the concept learning model of the hippocampus and the neocortex, to generate clusters and centroids for different scene categories. Test images depicting different scenes are classified by using their distance to the closest centroids (concepts). Modeling of RGB-D scenes as centroids not only leads to state-of-the-art classification performance on benchmark datasets (SUN RGB-D and NYU Depth V2), but also offers a method for inspecting and interpreting the space of centroids. Inspection of the centroids generated by our approach on RGB-D datasets leads us to propose a method for merging conceptually similar categories, resulting in improved accuracy for all approaches.

arXiv:1910.01967

Objective Human Affective Vocal Expression Detection and Automatic Classification with Stochastic Models and Learning Systems

This paper presents a widespread analysis of affective vocal expression classification systems. In this study, state-of-the-art acoustic features are compared to two novel affective vocal prints for the detection of emotional states: the Hilbert-Huang-Hurst Coefficients (HHHC) and the vector of index of non-stationarity (INS). HHHC is here proposed as a nonlinear vocal source feature vector that represents the affective states according to their effects on the speech production mechanism. Emotional states are highlighted by the empirical mode decomposition (EMD) based method, which exploits the non-stationarity of the affective acoustic variations. Hurst coefficients (closely related to the excitation source) are then estimated from the decomposition process to compose the feature vector. Additionally, the INS vector is introduced as dynamic information to the HHHC feature. The proposed features are evaluated in speech emotion classification experiments with three databases in German and English languages. Three state-of-the-art acoustic features are adopted as baseline. The $\alpha$-integrated Gaussian model ($\alpha$-GMM) is also introduced for the emotion representation and classification. Its performance is compared to competing stochastic and machine learning classifiers. Results demonstrate that HHHC leads to significant classification improvement when compared to the baseline acoustic features. Moreover, results also show that $\alpha$-GMM outperforms the competing classification methods. Finally, HHHC and INS are also evaluated as complementary features for the GeMAPS and eGeMAPS feature sets

arXiv:1911.00216

On Distributed Quantization for Classification

We consider the problem of distributed feature quantization, where the goal is to enable a pretrained classifier at a central node to carry out its classification on features that are gathered from distributed nodes through communication constrained channels. We propose the design of distributed quantization schemes specifically tailored to the classification task: unlike quantization schemes that help the central node reconstruct the original signal as accurately as possible, our focus is not reconstruction accuracy, but instead correct classification. Our work does not make any apriori distributional assumptions on the data, but instead uses training data for the quantizer design. Our main contributions include: we prove NP-hardness of finding optimal quantizers in the general case; we design an optimal scheme for a special case; we propose quantization algorithms, that leverage discrete neural representations and training data, and can be designed in polynomial-time for any number of features, any number of classes, and arbitrary division of features across the distributed nodes. We find that tailoring the quantizers to the classification task can offer significant savings: as compared to alternatives, we can achieve more than a factor of two reduction in terms of the number of bits communicated, for the same classification accuracy.

arXiv:1910.14056

Unsupervised Star Galaxy Classification with Cascade Variational Auto-Encoder

The increasing amount of data in astronomy provides great challenges for machine learning research. Previously, supervised learning methods achieved satisfactory recognition accuracy for the star-galaxy classification task, based on manually labeled data set. In this work, we propose a novel unsupervised approach for the star-galaxy recognition task, namely Cascade Variational Auto-Encoder (CasVAE). Our empirical results show our method outperforms the baseline model in both accuracy and stability.

arXiv:1910.14578

Very high resolution Airborne PolSAR Image Classification using Convolutional Neural Networks

In this work, we exploit convolutional neural networks (CNNs) for the classification of very high resolution (VHR) polarimetric SAR (PolSAR) data. Due to the significant appearance of heterogeneous textures within these data, not only polarimetric features but also structural tensors are exploited to feed CNN models. For deep networks, we use the SegNet model for semantic segmentation, which corresponds to pixelwise classification in remote sensing. Our experiments on the airborne F-SAR data show that for VHR PolSAR images, SegNet could provide high accuracy for the classification task; and introducing structural tensors together with polarimetric features as inputs could help the network to focus more on geometrical information to significantly improve the classification performance.

arXiv:1910.13830공식 코드 · Papers with Code

Extreme Classification in Log Memory using Count-Min Sketch: A Case Study of Amazon Search with 50M Products

In the last decade, it has been shown that many hard AI tasks, especially in NLP, can be naturally modeled as extreme classification problems leading to improved precision. However, such models are prohibitively expensive to train due to the memory blow-up in the last layer. For example, a reasonable softmax layer for the dataset of interest in this paper can easily reach well beyond 100 billion parameters (>400 GB memory). To alleviate this problem, we present Merged-Average Classifiers via Hashing (MACH), a generic K-classification algorithm where memory provably scales at O(logK) without any strong assumption on the classes. MACH is subtly a count-min sketch structure in disguise, which uses universal hashing to reduce classification with a large number of classes to few embarrassingly parallel and independent classification tasks with a small (constant) number of classes. MACH naturally provides a technique for zero communication model parallelism. We experiment with 6 datasets; some multiclass and some multilabel, and show consistent improvement over respective state-of-the-art baselines. In particular, we train an end-to-end deep classifier on a private product search dataset sampled from Amazon Search Engine with 70 million queries and 49.46 million products. MACH outperforms, by a significant margin,the state-of-the-art extreme classification models deployed on commercial search engines: Parabel and dense embedding models. Our largest model has 6.4 billion parameters and trains in less than 35 hours on a single p3.16x machine. Our training times are 7-10x faster, and our memory footprints are 2-4x smaller than the best baselines. This training time is also significantly lower than the one reported by Google's mixture of experts (MoE) language model on a comparable model size and hardware.

arXiv:1910.13631

When does Diversity Help Generalization in Classification Ensembles?

Ensembles, as a widely used and effective technique in the machine learning community, succeed within a key element -- "diversity." The relationship between diversity and generalization, unfortunately, is not entirely understood and remains an open research issue. To reveal the effect of diversity on the generalization of classification ensembles, we investigate three issues on diversity, i.e., the measurement of diversity, the relationship between the proposed diversity and the generalization error, and the utilization of this relationship for ensemble pruning. In the diversity measurement, we measure diversity by error decomposition inspired by regression ensembles, which decomposes the error of classification ensembles into accuracy and diversity. Then we formulate the relationship between the measured diversity and ensemble performance through the theorem of margin and generalization and observe that the generalization error is reduced effectively only when the measured diversity is increased in a few specific ranges, while in other ranges larger diversity is less beneficial to increasing the generalization of an ensemble. Besides, we propose two pruning methods based on diversity management to utilize this relationship, which could increase diversity appropriately and shrink the size of the ensemble without much-decreasing performance. Empirical results validate the reasonableness of the proposed relationship between diversity and ensemble generalization error and the effectiveness of the proposed pruning methods.

arXiv:1910.13328

Weakly Supervised Prostate TMA Classification via Graph Convolutional Networks

Histology-based grade classification is clinically important for many cancer types in stratifying patients distinct treatment groups. In prostate cancer, the Gleason score is a grading system used to measure the aggressiveness of prostate cancer from the spatial organization of cells and the distribution of glands. However, the subjective interpretation of Gleason score often suffers from large interobserver and intraobserver variability. Previous work in deep learning-based objective Gleason grading requires manual pixel-level annotation. In this work, we propose a weakly-supervised approach for grade classification in tissue micro-arrays (TMA) using graph convolutional networks (GCNs), in which we model the spatial organization of cells as a graph to better capture the proliferation and community structure of tumor cells. As node-level features in our graph representation, we learn the morphometry of each cell using a contrastive predictive coding (CPC)-based self-supervised approach. We demonstrate that on a five-fold cross validation our method can achieve $0.9659\pm0.0096$ AUC using only TMA-level labels. Our method demonstrates a 39.80\% improvement over standard GCNs with texture features and a 29.27% improvement over GCNs with VGG19 features. Our proposed pipeline can be used to objectively stratify low and high risk cases, reducing inter- and intra-observer variability and pathologist workload.

arXiv:1910.13081공식 코드 · Papers with Code

Classification Calibration for Long-tail Instance Segmentation

Remarkable progress has been made in object instance detection and segmentation in recent years. However, existing state-of-the-art methods are mostly evaluated with fairly balanced and class-limited benchmarks, such as Microsoft COCO dataset [8]. In this report, we investigate the performance drop phenomenon of state-of-the-art two-stage instance segmentation models when processing extreme long-tail training data based on the LVIS [5] dataset, and find a major cause is the inaccurate classification of object proposals. Based on this observation, we propose to calibrate the prediction of classification head to improve recognition performance for the tail classes. Without much additional cost and modification of the detection model architecture, our calibration method improves the performance of the baseline by a large margin on the tail classes. Codes will be available. Importantly, after the submission, we find significant improvement can be further achieved by modifying the calibration head, which we will update later.

arXiv:1910.12861

Deep Learning for Hyperspectral Image Classification: An Overview

Hyperspectral image (HSI) classification has become a hot topic in the field of remote sensing. In general, the complex characteristics of hyperspectral data make the accurate classification of such data challenging for traditional machine learning methods. In addition, hyperspectral imaging often deals with an inherently nonlinear relation between the captured spectral information and the corresponding materials. In recent years, deep learning has been recognized as a powerful feature-extraction tool to effectively address nonlinear problems and widely used in a number of image processing tasks. Motivated by those successful applications, deep learning has also been introduced to classify HSIs and demonstrated good performance. This survey paper presents a systematic review of deep learning-based HSI classification literatures and compares several strategies for this topic. Specifically, we first summarize the main challenges of HSI classification which cannot be effectively overcome by traditional machine learning methods, and also introduce the advantages of deep learning to handle these problems. Then, we build a framework which divides the corresponding works into spectral-feature networks, spatial-feature networks, and spectral-spatial-feature networks to systematically review the recent achievements in deep learning-based HSI classification. In addition, considering the fact that available training samples in the remote sensing field are usually very limited and training deep networks require a large number of samples, we include some strategies to improve classification performance, which can provide some guidelines for future studies on this topic. Finally, several representative deep learning-based classification methods are conducted on real HSIs in our experiments.

arXiv:1910.13425

Weakly-Supervised Deep Learning for Domain Invariant Sentiment Classification

The task of learning a sentiment classification model that adapts well to any target domain, different from the source domain, is a challenging problem. Majority of the existing approaches focus on learning a common representation by leveraging both source and target data during training. In this paper, we introduce a two-stage training procedure that leverages weakly supervised datasets for developing simple lift-and-shift-based predictive models without being exposed to the target domain during the training phase. Experimental results show that transfer with weak supervision from a source domain to various target domains provides performance very close to that obtained via supervised training on the target domain itself.