PyTorchKR Papers

최신 논문

총 4,618편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1408.4692

Seeing through bag-of-visual-word glasses: towards understanding quantization effects in feature extraction methods

Vector-quantized local features frequently used in bag-of-visual-words approaches are the backbone of popular visual recognition systems due to both their simplicity and their performance. Despite their success, bag-of-words-histograms basically contain low-level image statistics (e.g., number of edges of different orientations). The question remains how much visual information is "lost in quantization" when mapping visual features to code words? To answer this question, we present an in-depth analysis of the effect of local feature quantization on human recognition performance. Our analysis is based on recovering the visual information by inverting quantized local features and presenting these visualizations with different codebook sizes to human observers. Although feature inversion techniques are around for quite a while, to the best of our knowledge, our technique is the first visualizing especially the effect of feature quantization. Thereby, we are now able to compare single steps in common image classification pipelines to human counterparts.

arXiv:1409.8211

Efficient multivariate sequence classification

Kernel-based approaches for sequence classification have been successfully applied to a variety of domains, including the text categorization, image classification, speech analysis, biological sequence analysis, time series and music classification, where they show some of the most accurate results. Typical kernel functions for sequences in these domains (e.g., bag-of-words, mismatch, or subsequence kernels) are restricted to {\em discrete univariate} (i.e. one-dimensional) string data, such as sequences of words in the text analysis, codeword sequences in the image analysis, or nucleotide or amino acid sequences in the DNA and protein sequence analysis. However, original sequence data are often of real-valued multivariate nature, i.e. are not univariate and discrete as required by typical $k$-mer based sequence kernel functions. In this work, we consider the problem of the {\em multivariate} sequence classification such as classification of multivariate music sequences, or multidimensional protein sequence representations. To this end, we extend {\em univariate} kernel functions typically used in sequence analysis and propose efficient {\em multivariate} similarity kernel method (MVDFQ-SK) based on (1) a direct feature quantization (DFQ) of each sequence dimension in the original {\em real-valued} multivariate sequences and (2) applying novel multivariate discrete kernel measures on these multivariate discrete DFQ sequence representations to more accurately capture similarity relationships among sequences and improve classification performance. Experiments using the proposed MVDFQ-SK kernel method show excellent classification performance on three challenging music classification tasks as well as protein sequence classification with significant 25-40% improvements over univariate kernel methods and existing state-of-the-art sequence classification methods.

arXiv:1411.2795

Speaker Identification From Youtube Obtained Data

An efficient, and intuitive algorithm is presented for the identification of speakers from a long dataset (like YouTube long discussion, Cocktail party recorded audio or video).The goal of automatic speaker identification is to identify the number of different speakers and prepare a model for that speaker by extraction, characterization and speaker-specific information contained in the speech signal. It has many diverse application specially in the field of Surveillance, Immigrations at Airport, cyber security, transcription in multi-source of similar sound source, where it is difficult to assign transcription arbitrary. The most commonly speech parametrization used in speaker verification, K-mean, cepstral analysis, is detailed. Gaussian mixture modeling, which is the speaker modeling technique is then explained. Gaussian mixture models (GMM), perhaps the most robust machine learning algorithm has been introduced examine and judge carefully speaker identification in text independent. The application or employment of Gaussian mixture models for monitoring & Analysing speaker identity is encouraged by the familiarity, awareness, or understanding gained through experience that Gaussian spectrum depict the characteristics of speaker's spectral conformational pattern and remarkable ability of GMM to construct capricious densities after that we illustrate 'Expectation maximization' an iterative algorithm which takes some arbitrary value in initial estimation and carry on the iterative process until the convergence of value is observed,so by doing various number of experiments we are able to obtain 79 ~ 82% of identification rate using Vector quantization and 85 ~ 92.6% of identification rate using GMM modeling by Expectation maximization parameter estimation depending on variation of parameter.

arXiv:1411.2173공식 코드 · Papers with Code

Stacked Quantizers for Compositional Vector Compression

Recently, Babenko and Lempitsky introduced Additive Quantization (AQ), a generalization of Product Quantization (PQ) where a non-independent set of codebooks is used to compress vectors into small binary codes. Unfortunately, under this scheme encoding cannot be done independently in each codebook, and optimal encoding is an NP-hard problem. In this paper, we observe that PQ and AQ are both compositional quantizers that lie on the extremes of the codebook dependence-independence assumption, and explore an intermediate approach that exploits a hierarchical structure in the codebooks. This results in a method that achieves quantization error on par with or lower than AQ, while being several orders of magnitude faster. We perform a complexity analysis of PQ, AQ and our method, and evaluate our approach on standard benchmarks of SIFT and GIST descriptors, as well as on new datasets of features obtained from state-of-the-art convolutional neural networks.

arXiv:1407.4764

Efficient On-the-fly Category Retrieval using ConvNets and GPUs

We investigate the gains in precision and speed, that can be obtained by using Convolutional Networks (ConvNets) for on-the-fly retrieval - where classifiers are learnt at run time for a textual query from downloaded images, and used to rank large image or video datasets. We make three contributions: (i) we present an evaluation of state-of-the-art image representations for object category retrieval over standard benchmark datasets containing 1M+ images; (ii) we show that ConvNets can be used to obtain features which are incredibly performant, and yet much lower dimensional than previous state-of-the-art image representations, and that their dimensionality can be reduced further without loss in performance by compression using product quantization or binarization. Consequently, features with the state-of-the-art performance on large-scale datasets of millions of images can fit in the memory of even a commodity GPU card; (iii) we show that an SVM classifier can be learnt within a ConvNet framework on a GPU in parallel with downloading the new training images, allowing for a continuous refinement of the model as more images become available, and simultaneous training and ranking. The outcome is an on-the-fly system that significantly outperforms its predecessors in terms of: precision of retrieval, memory requirements, and speed, facilitating accurate on-the-fly learning and ranking in under a second on a single GPU.

arXiv:1412.0826

Hashing on Nonlinear Manifolds

Learning based hashing methods have attracted considerable attention due to their ability to greatly increase the scale at which existing algorithms may operate. Most of these methods are designed to generate binary codes preserving the Euclidean similarity in the original space. Manifold learning techniques, in contrast, are better able to model the intrinsic structure embedded in the original high-dimensional data. The complexities of these models, and the problems with out-of-sample data, have previously rendered them unsuitable for application to large-scale embedding, however. In this work, how to learn compact binary embeddings on their intrinsic manifolds is considered. In order to address the above-mentioned difficulties, an efficient, inductive solution to the out-of-sample data problem, and a process by which non-parametric manifold learning may be used as the basis of a hashing method is proposed. The proposed approach thus allows the development of a range of new hashing techniques exploiting the flexibility of the wide variety of manifold learning approaches available. It is particularly shown that hashing on the basis of t-SNE outperforms state-of-the-art hashing methods on large-scale benchmark datasets, and is very effective for image classification with very short code lengths. The proposed hashing framework is shown to be easily improved, for example, by minimizing the quantization error with learned orthogonal rotations. In addition, a supervised inductive manifold hashing framework is developed by incorporating the label information, which is shown to greatly advance the semantic retrieval performance.

arXiv:1412.6115

Compressing Deep Convolutional Networks using Vector Quantization

Deep convolutional neural networks (CNN) has become the most promising method for object recognition, repeatedly demonstrating record breaking results for image classification and object detection in recent years. However, a very deep CNN generally involves many layers with millions of parameters, making the storage of the network model to be extremely large. This prohibits the usage of deep CNNs on resource limited hardware, especially cell phones or other embedded devices. In this paper, we tackle this model storage issue by investigating information theoretical vector quantization methods for compressing the parameters of CNNs. In particular, we have found in terms of compressing the most storage demanding dense connected layers, vector quantization methods have a clear gain over existing matrix factorization methods. Simply applying k-means clustering to the weights or conducting product quantization can lead to a very good balance between model size and recognition accuracy. For the 1000-category classification task in the ImageNet challenge, we are able to achieve 16-24 times compression of the network with only 1% loss of classification accuracy using the state-of-the-art CNN.

arXiv:1501.00105

Face recognition using color local binary pattern from mutually independent color channels

In this paper, a high performance face recognition system based on local binary pattern (LBP) using the probability distribution functions (PDF) of pixels in different mutually independent color channels which are robust to frontal homogenous illumination and planer rotation is proposed. The illumination of faces is enhanced by using the state-of-the-art technique which is using discrete wavelet transform (DWT) and singular value decomposition (SVD). After equalization, face images are segmented by use of local Successive Mean Quantization Transform (SMQT) followed by skin color based face detection system. Kullback-Leibler Distance (KLD) between the concatenated PDFs of a given face obtained by LBP and the concatenated PDFs of each face in the database is used as a metric in the recognition process. Various decision fusion techniques have been used in order to improve the recognition rate. The proposed system has been tested on the FERET, HP, and Bosphorus face databases. The proposed system is compared with conventional and thestate-of-the-art techniques. The recognition rates obtained using FVF approach for FERET database is 99.78% compared with 79.60% and 68.80% for conventional gray scale LBP and Principle Component Analysis (PCA) based face recognition techniques respectively.

arXiv:1501.07681

Vector Quantization by Minimizing Kullback-Leibler Divergence

This paper proposes a new method for vector quantization by minimizing the Kullback-Leibler Divergence between the class label distributions over the quantization inputs, which are original vectors, and the output, which is the quantization subsets of the vector set. In this way, the vector quantization output can keep as much information of the class label as possible. An objective function is constructed and we also developed an iterative algorithm to minimize it. The new method is evaluated on bag-of-features based image classification problem.

arXiv:1501.07422

Pairwise Rotation Hashing for High-dimensional Features

Binary Hashing is widely used for effective approximate nearest neighbors search. Even though various binary hashing methods have been proposed, very few methods are feasible for extremely high-dimensional features often used in visual tasks today. We propose a novel highly sparse linear hashing method based on pairwise rotations. The encoding cost of the proposed algorithm is $\mathrm{O}(n \log n)$ for n-dimensional features, whereas that of the existing state-of-the-art method is typically $\mathrm{O}(n^2)$. The proposed method is also remarkably faster in the learning phase. Along with the efficiency, the retrieval accuracy is comparable to or slightly outperforming the state-of-the-art. Pairwise rotations used in our method are formulated from an analytical study of the trade-off relationship between quantization error and entropy of binary codes. Although these hashing criteria are widely used in previous researches, its analytical behavior is rarely studied. All building blocks of our algorithm are based on the analytical solution, and it thus provides a fairly simple and efficient procedure.

arXiv:1412.1945

Background Modelling using Octree Color Quantization

By assuming that the most frequently occuring color in a video or a region of a video I propose a new algorithm for detecting foreground objects in a video. The process of detecting the foreground objects is complicated because of the fact that there may be swaying trees, objects of the background being moved around or lighting changes in the video. To deal with such complexities many have come up with solutions which heavily rely on expensive floating point operations. In this paper I used a data structure called Octree which is implemented only using binary operations. Traditionally octrees were used for color quantization but here in this paper I used it as a data structure to store the most frequently occuring colors in a video as well. For each of the starting few video frames, I constructed a Octree using all the colors of that frame. Next I pruned all the trees by removing nodes below a certain height and gave the leaf nodes a color which is dependant on the topological path from that node to its parent. Hence any two leaf nodes in two different octrees with the same topological path from themselves to the root will represent the same color. Next I merged all these individual trees into a single tree retaining only those nodes whose topological path to itself from the root is most common among all the trees. The colors represented by the leaf nodes in the resultant tree will be the most frequently occuring colors in the starting video frames of the video. Hence any color of an incomming frame that is not close to any of the colors represented by the leaf node of the merged tree can be regarded as belonging to a foreground object. As an Octree is constructed using only binary operations, it is very fast compared to other leading algorithms.

arXiv:1503.08322

Some Further Evidence about Magnification and Shape in Neural Gas

Neural gas (NG) is a robust vector quantization algorithm with a well-known mathematical model. According to this, the neural gas samples the underlying data distribution following a power law with a magnification exponent that depends on data dimensionality only. The effects of shape in the input data distribution, however, are not entirely covered by the NG model above, due to the technical difficulties involved. The experimental work described here shows that shape is indeed relevant in determining the overall NG behavior; in particular, some experiments reveal richer and complex behaviors induced by shape that cannot be explained by the power law alone. Although a more comprehensive analytical model remains to be defined, the evidence collected in these experiments suggests that the NG algorithm has an interesting potential for detecting complex shapes in noisy datasets.

arXiv:1502.05313

Variational Optimization of Annealing Schedules

Annealed importance sampling (AIS) is a common algorithm to estimate partition functions of useful stochastic models. One important problem for obtaining accurate AIS estimates is the selection of an annealing schedule. Conventionally, an annealing schedule is often determined heuristically or is simply set as a linearly increasing sequence. In this paper, we propose an algorithm for the optimal schedule by deriving a functional that dominates the AIS estimation error and by numerically minimizing this functional. We experimentally demonstrate that the proposed algorithm mostly outperforms conventional scheduling schemes with large quantization numbers.

arXiv:1504.03410

Simultaneous Feature Learning and Hash Coding with Deep Neural Networks

Similarity-preserving hashing is a widely-used method for nearest neighbour search in large-scale image retrieval tasks. For most existing hashing methods, an image is first encoded as a vector of hand-engineering visual features, followed by another separate projection or quantization step that generates binary codes. However, such visual feature vectors may not be optimally compatible with the coding process, thus producing sub-optimal hashing codes. In this paper, we propose a deep architecture for supervised hashing, in which images are mapped into binary codes via carefully designed deep neural networks. The pipeline of the proposed deep architecture consists of three building blocks: 1) a sub-network with a stack of convolution layers to produce the effective intermediate image features; 2) a divide-and-encode module to divide the intermediate image features into multiple branches, each encoded into one hash bit; and 3) a triplet ranking loss designed to characterize that one image is more similar to the second image than to the third one. Extensive evaluations on several benchmark image datasets show that the proposed simultaneous feature learning and hash coding pipeline brings substantial improvements over other state-of-the-art supervised or unsupervised hashing methods.

arXiv:1506.01092

Bilinear Random Projections for Locality-Sensitive Binary Codes

Locality-sensitive hashing (LSH) is a popular data-independent indexing method for approximate similarity search, where random projections followed by quantization hash the points from the database so as to ensure that the probability of collision is much higher for objects that are close to each other than for those that are far apart. Most of high-dimensional visual descriptors for images exhibit a natural matrix structure. When visual descriptors are represented by high-dimensional feature vectors and long binary codes are assigned, a random projection matrix requires expensive complexities in both space and time. In this paper we analyze a bilinear random projection method where feature matrices are transformed to binary codes by two smaller random projection matrices. We base our theoretical analysis on extending Raginsky and Lazebnik's result where random Fourier features are composed with random binary quantizers to form locality sensitive binary codes. To this end, we answer the following two questions: (1) whether a bilinear random projection also yields similarity-preserving binary codes; (2) whether a bilinear random projection yields performance gain or loss, compared to a large linear projection. Regarding the first question, we present upper and lower bounds on the expected Hamming distance between binary codes produced by bilinear random projections. In regards to the second question, we analyze the upper and lower bounds on covariance between two bits of binary codes, showing that the correlation between two bits is small. Numerical experiments on MNIST and Flickr45K datasets confirm the validity of our method.

arXiv:1506.02717

An Improved BKW Algorithm for LWE with Applications to Cryptography and Lattices

In this paper, we study the Learning With Errors problem and its binary variant, where secrets and errors are binary or taken in a small interval. We introduce a new variant of the Blum, Kalai and Wasserman algorithm, relying on a quantization step that generalizes and fine-tunes modulus switching. In general this new technique yields a significant gain in the constant in front of the exponent in the overall complexity. We illustrate this by solving p within half a day a LWE instance with dimension n = 128, modulus $q = n^2$, Gaussian noise $\alpha = 1/(\sqrt{n/\pi} \log^2 n)$ and binary secret, using $2^{28}$ samples, while the previous best result based on BKW claims a time complexity of $2^{74}$ with $2^{60}$ samples for the same parameters. We then introduce variants of BDD, GapSVP and UniqueSVP, where the target point is required to lie in the fundamental parallelepiped, and show how the previous algorithm is able to solve these variants in subexponential time. Moreover, we also show how the previous algorithm can be used to solve the BinaryLWE problem with n samples in subexponential time $2^{(\ln 2/2+o(1))n/\log \log n}$. This analysis does not require any heuristic assumption, contrary to other algebraic approaches; instead, it uses a variant of an idea by Lyubashevsky to generate many samples from a small number of samples. This makes it possible to asymptotically and heuristically break the NTRU cryptosystem in subexponential time (without contradicting its security assumption). We are also able to solve subset sum problems in subexponential time for density $o(1)$, which is of independent interest: for such density, the previous best algorithm requires exponential time. As a direct application, we can solve in subexponential time the parameters of a cryptosystem based on this problem proposed at TCC 2010.

arXiv:1507.01442

Learning Better Encoding for Approximate Nearest Neighbor Search with Dictionary Annealing

We introduce a novel dictionary optimization method for high-dimensional vector quantization employed in approximate nearest neighbor (ANN) search. Vector quantization methods first seek a series of dictionaries, then approximate each vector by a sum of elements selected from these dictionaries. An optimal series of dictionaries should be mutually independent, and each dictionary should generate a balanced encoding for the target dataset. Existing methods did not explicitly consider this. To achieve these goals along with minimizing the quantization error (residue), we propose a novel dictionary optimization method called \emph{Dictionary Annealing} that alternatively "heats up" a single dictionary by generating an intermediate dataset with residual vectors, "cools down" the dictionary by fitting the intermediate dataset, then extracts the new residual vectors for the next iteration. Better codes can be learned by DA for the ANN search tasks. DA is easily implemented on GPU to utilize the latest computing technology, and can easily extended to an online dictionary learning scheme. We show by experiments that our optimized dictionaries substantially reduce the overall quantization error. Jointly used with residual vector quantization, our optimized dictionaries lead to a better approximate nearest neighbor search performance compared to the state-of-the-art methods.

arXiv:1507.04437

A Deep Hashing Learning Network

Hashing-based methods seek compact and efficient binary codes that preserve the neighborhood structure in the original data space. For most existing hashing methods, an image is first encoded as a vector of hand-crafted visual feature, followed by a hash projection and quantization step to get the compact binary vector. Most of the hand-crafted features just encode the low-level information of the input, the feature may not preserve the semantic similarities of images pairs. Meanwhile, the hashing function learning process is independent with the feature representation, so the feature may not be optimal for the hashing projection. In this paper, we propose a supervised hashing method based on a well designed deep convolutional neural network, which tries to learn hashing code and compact representations of data simultaneously. The proposed model learn the binary codes by adding a compact sigmoid layer before the loss layer. Experiments on several image data sets show that the proposed model outperforms other state-of-the-art methods.

arXiv:1509.01469

Quantization based Fast Inner Product Search

We propose a quantization based approach for fast approximate Maximum Inner Product Search (MIPS). Each database vector is quantized in multiple subspaces via a set of codebooks, learned directly by minimizing the inner product quantization error. Then, the inner product of a query to a database vector is approximated as the sum of inner products with the subspace quantizers. Different from recently proposed LSH approaches to MIPS, the database vectors and queries do not need to be augmented in a higher dimensional feature space. We also provide a theoretical analysis of the proposed approach, consisting of the concentration results under mild assumptions. Furthermore, if a small sample of example queries is given at the training time, we propose a modified codebook learning procedure which further improves the accuracy. Experimental results on a variety of datasets including those arising from deep neural networks show that the proposed approach significantly outperforms the existing state-of-the-art.

arXiv:1509.05195

Improved Residual Vector Quantization for High-dimensional Approximate Nearest Neighbor Search

Quantization methods have been introduced to perform large scale approximate nearest search tasks. Residual Vector Quantization (RVQ) is one of the effective quantization methods. RVQ uses a multi-stage codebook learning scheme to lower the quantization error stage by stage. However, there are two major limitations for RVQ when applied to on high-dimensional approximate nearest neighbor search: 1. The performance gain diminishes quickly with added stages. 2. Encoding a vector with RVQ is actually NP-hard. In this paper, we propose an improved residual vector quantization (IRVQ) method, our IRVQ learns codebook with a hybrid method of subspace clustering and warm-started k-means on each stage to prevent performance gain from dropping, and uses a multi-path encoding scheme to encode a vector with lower distortion. Experimental results on the benchmark datasets show that our method gives substantially improves RVQ and delivers better performance compared to the state-of-the-art.