PyTorchKR Papers

최신 논문

총 4,618편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1711.01514

Distribution-Preserving k-Anonymity

Preserving the privacy of individuals by protecting their sensitive attributes is an important consideration during microdata release. However, it is equally important to preserve the quality or utility of the data for at least some targeted workloads. We propose a novel framework for privacy preservation based on the k-anonymity model that is ideally suited for workloads that require preserving the probability distribution of the quasi-identifier variables in the data. Our framework combines the principles of distribution-preserving quantization and k-member clustering, and we specialize it to two variants that respectively use intra-cluster and Gaussian dithering of cluster centers to achieve distribution preservation. We perform theoretical analysis of the proposed schemes in terms of distribution preservation, and describe their utility in workloads such as covariate shift and transfer learning where such a property is necessary. Using extensive experiments on real-world Medical Expenditure Panel Survey data, we demonstrate the merits of our algorithms over standard k-anonymization for a hallmark health care application where an insurance company wishes to understand the risk in entering a new market. Furthermore, by empirically quantifying the reidentification risk, we also show that the proposed approaches indeed maintain k-anonymity.

arXiv:1711.02604

Unbounded cache model for online language modeling with open vocabulary

Recently, continuous cache models were proposed as extensions to recurrent neural network language models, to adapt their predictions to local changes in the data distribution. These models only capture the local context, of up to a few thousands tokens. In this paper, we propose an extension of continuous cache models, which can scale to larger contexts. In particular, we use a large scale non-parametric memory component that stores all the hidden activations seen in the past. We leverage recent advances in approximate nearest neighbor search and quantization algorithms to store millions of representations while searching them efficiently. We conduct extensive experiments showing that our approach significantly improves the perplexity of pre-trained language models on new distributions, and can scale efficiently to much larger contexts than previously proposed local cache models.

arXiv:1711.03712

Quantized Memory-Augmented Neural Networks

Memory-augmented neural networks (MANNs) refer to a class of neural network models equipped with external memory (such as neural Turing machines and memory networks). These neural networks outperform conventional recurrent neural networks (RNNs) in terms of learning long-term dependency, allowing them to solve intriguing AI tasks that would otherwise be hard to address. This paper concerns the problem of quantizing MANNs. Quantization is known to be effective when we deploy deep models on embedded systems with limited resources. Furthermore, quantization can substantially reduce the energy consumption of the inference procedure. These benefits justify recent developments of quantized multi layer perceptrons, convolutional networks, and RNNs. However, no prior work has reported the successful quantization of MANNs. The in-depth analysis presented here reveals various challenges that do not appear in the quantization of the other networks. Without addressing them properly, quantized MANNs would normally suffer from excessive quantization error which leads to degraded performance. In this paper, we identify memory addressing (specifically, content-based addressing) as the main reason for the performance degradation and propose a robust quantization method for MANNs to address the challenge. In our experiments, we achieved a computation-energy gain of 22x with 8-bit fixed-point and binary quantization compared to the floating-point implementation. Measured on the bAbI dataset, the resulting model, named the quantized MANN (Q-MANN), improved the error rate by 46% and 30% with 8-bit fixed-point and binary quantization, respectively, compared to the MANN quantized using conventional techniques.

arXiv:1711.04713

ADaPTION: Toolbox and Benchmark for Training Convolutional Neural Networks with Reduced Numerical Precision Weights and Activation

Deep Neural Networks (DNNs) and Convolutional Neural Networks (CNNs) are useful for many practical tasks in machine learning. Synaptic weights, as well as neuron activation functions within the deep network are typically stored with high-precision formats, e.g. 32 bit floating point. However, since storage capacity is limited and each memory access consumes power, both storage capacity and memory access are two crucial factors in these networks. Here we present a method and present the ADaPTION toolbox to extend the popular deep learning library Caffe to support training of deep CNNs with reduced numerical precision of weights and activations using fixed point notation. ADaPTION includes tools to measure the dynamic range of weights and activations. Using the ADaPTION tools, we quantized several CNNs including VGG16 down to 16-bit weights and activations with only 0.8% drop in Top-1 accuracy. The quantization, especially of the activations, leads to increase of up to 50% of sparsity especially in early and intermediate layers, which we exploit to skip multiplications with zero, thus performing faster and computationally cheaper inference.

arXiv:1711.04686

Weightless: Lossy Weight Encoding For Deep Neural Network Compression

The large memory requirements of deep neural networks limit their deployment and adoption on many devices. Model compression methods effectively reduce the memory requirements of these models, usually through applying transformations such as weight pruning or quantization. In this paper, we present a novel scheme for lossy weight encoding which complements conventional compression techniques. The encoding is based on the Bloomier filter, a probabilistic data structure that can save space at the cost of introducing random errors. Leveraging the ability of neural networks to tolerate these imperfections and by re-training around the errors, the proposed technique, Weightless, can compress DNN weights by up to 496x with the same model accuracy. This results in up to a 1.51x improvement over the state-of-the-art.

arXiv:1612.01543

Towards the Limit of Network Quantization

Network quantization is one of network compression techniques to reduce the redundancy of deep neural networks. It reduces the number of distinct network parameter values by quantization in order to save the storage for them. In this paper, we design network quantization schemes that minimize the performance loss due to quantization given a compression ratio constraint. We analyze the quantitative relation of quantization errors to the neural network loss function and identify that the Hessian-weighted distortion measure is locally the right objective function for the optimization of network quantization. As a result, Hessian-weighted k-means clustering is proposed for clustering network parameters to quantize. When optimal variable-length binary codes, e.g., Huffman codes, are employed for further compression, we derive that the network quantization problem can be related to the entropy-constrained scalar quantization (ECSQ) problem in information theory and consequently propose two solutions of ECSQ for network quantization, i.e., uniform quantization and an iterative solution similar to Lloyd's algorithm. Finally, using the simple uniform quantization followed by Huffman coding, we show from our experiments that the compression ratios of 51.25, 22.17 and 40.65 are achievable for LeNet, 32-layer ResNet and AlexNet, respectively.

arXiv:1711.05738

The Neural Network Pushdown Automaton: Model, Stack and Learning Simulations

In order for neural networks to learn complex languages or grammars, they must have sufficient computational power or resources to recognize or generate such languages. Though many approaches have been discussed, one ob- vious approach to enhancing the processing power of a recurrent neural network is to couple it with an external stack memory - in effect creating a neural network pushdown automata (NNPDA). This paper discusses in detail this NNPDA - its construction, how it can be trained and how useful symbolic information can be extracted from the trained network. In order to couple the external stack to the neural network, an optimization method is developed which uses an error function that connects the learning of the state automaton of the neural network to the learning of the operation of the external stack. To minimize the error function using gradient descent learning, an analog stack is designed such that the action and storage of information in the stack are continuous. One interpretation of a continuous stack is the probabilistic storage of and action on data. After training on sample strings of an unknown source grammar, a quantization procedure extracts from the analog stack and neural network a discrete pushdown automata (PDA). Simulations show that in learning deterministic context-free grammars - the balanced parenthesis language, 1*n0*n, and the deterministic Palindrome - the extracted PDA is correct in the sense that it can correctly recognize unseen strings of arbitrary length. In addition, the extracted PDAs can be shown to be identical or equivalent to the PDAs of the source grammars which were used to generate the training strings.

arXiv:1711.00205

Towards Effective Low-bitwidth Convolutional Neural Networks

This paper tackles the problem of training a deep convolutional neural network with both low-precision weights and low-bitwidth activations. Optimizing a low-precision network is very challenging since the training process can easily get trapped in a poor local minima, which results in substantial accuracy loss. To mitigate this problem, we propose three simple-yet-effective approaches to improve the network training. First, we propose to use a two-stage optimization strategy to progressively find good local minima. Specifically, we propose to first optimize a net with quantized weights and then quantized activations. This is in contrast to the traditional methods which optimize them simultaneously. Second, following a similar spirit of the first method, we propose another progressive optimization approach which progressively decreases the bit-width from high-precision to low-precision during the course of training. Third, we adopt a novel learning scheme to jointly train a full-precision model alongside the low-precision one. By doing so, the full-precision model provides hints to guide the low-precision model training. Extensive experiments on various datasets ( i.e., CIFAR-100 and ImageNet) show the effectiveness of the proposed methods. To highlight, using our methods to train a 4-bit precision network leads to no performance decrease in comparison with its full-precision counterpart with standard network architectures ( i.e., AlexNet and ResNet-50).

arXiv:1711.01068

Compressing Word Embeddings via Deep Compositional Code Learning

Natural language processing (NLP) models often require a massive number of parameters for word embeddings, resulting in a large storage or memory footprint. Deploying neural NLP models to mobile devices requires compressing the word embeddings without any significant sacrifices in performance. For this purpose, we propose to construct the embeddings with few basis vectors. For each word, the composition of basis vectors is determined by a hash code. To maximize the compression rate, we adopt the multi-codebook quantization approach instead of binary coding scheme. Each code is composed of multiple discrete numbers, such as (3, 2, 1, 8), where the value of each component is limited to a fixed range. We propose to directly learn the discrete codes in an end-to-end neural network by applying the Gumbel-softmax trick. Experiments show the compression rate achieves 98% in a sentiment analysis task and 94% ~ 99% in machine translation tasks without performance loss. In both tasks, the proposed method can improve the model performance by slightly lowering the compression rate. Compared to other approaches such as character-level segmentation, the proposed method is language-independent and does not require modifications to the network architecture.

arXiv:1711.08589

End-to-End Supervised Product Quantization for Image Search and Retrieval

Product Quantization, a dictionary based hashing method, is one of the leading unsupervised hashing techniques. While it ignores the labels, it harnesses the features to construct look up tables that can approximate the feature space. In recent years, several works have achieved state of the art results on hashing benchmarks by learning binary representations in a supervised manner. This work presents Deep Product Quantization (DPQ), a technique that leads to more accurate retrieval and classification than the latest state of the art methods, while having similar computational complexity and memory footprint as the Product Quantization method. To our knowledge, this is the first work to introduce a dictionary-based representation that is inspired by Product Quantization and which is learned end-to-end, and thus benefits from the supervised signal. DPQ explicitly learns soft and hard representations to enable an efficient and accurate asymmetric search, by using a straight-through estimator. Our method obtains state of the art results on an extensive array of retrieval and classification experiments.

arXiv:1711.09156

Warped-Linear Models for Time Series Classification

This article proposes and studies warped-linear models for time series classification. The proposed models are time-warp invariant analogues of linear models. Their construction is in line with time series averaging and extensions of k-means and learning vector quantization to dynamic time warping (DTW) spaces. The main theoretical result is that warped-linear models correspond to polyhedral classifiers in Euclidean spaces. This result simplifies the analysis of time-warp invariant models by reducing to max-linear functions. We exploit this relationship and derive solutions to the label-dependency problem and the problem of learning warped-linear models. Empirical results on time series classification suggest that warped-linear functions better trade solution quality against computation time than nearest-neighbor and prototype-based methods.

arXiv:1711.06809

A Genetic Algorithm Approach for ImageRepresentation Learning through Color Quantization

Over the last decades, hand-crafted feature extractors have been used to encode image visual properties into feature vectors. Recently, data-driven feature learning approaches have been successfully explored as alternatives for producing more representative visual features. In this work, we combine both research venues, focusing on the color quantization problem. We propose two data-driven approaches to learn image representations through the search for optimized quantization schemes, which lead to more effective feature extraction algorithms and compact representations. Our strategy employs Genetic Algorithm, a soft-computing apparatus successfully utilized in Information-retrieval-related optimization problems. We hypothesize that changing the quantization affects the quality of image description approaches, leading to effective and efficient representations. We evaluate our approaches in content-based image retrieval tasks, considering eight well-known datasets with different visual properties. Results indicate that the approach focused on representation effectiveness outperformed baselines in all tested scenarios. The other approach, which also considers the size of created representations, produced competitive results keeping or even reducing the dimensionality of feature vectors up to 25%.

arXiv:1711.09256

Expectation maximization transfer learning and its application for bionic hand prostheses

Machine learning models in practical settings are typically confronted with changes to the distribution of the incoming data. Such changes can severely affect the model performance, leading for example to misclassifications of data. This is particularly apparent in the domain of bionic hand prostheses, where machine learning models promise faster and more intuitive user interfaces, but are hindered by their lack of robustness to everyday disturbances, such as electrode shifts. One way to address changes in the data distribution is transfer learning, that is, to transfer the disturbed data to a space where the original model is applicable again. In this contribution, we propose a novel expectation maximization algorithm to learn linear transformations that maximize the likelihood of disturbed data after the transformation. We also show that this approach generalizes to discriminative models, in particular learning vector quantization models. In our evaluation on data from the bionic prostheses domain we demonstrate that our approach can learn a transformation which improves classification accuracy significantly and outperforms all tested baselines, if few data or few classes are available in the target domain.

arXiv:1709.09883공식 코드 · Papers with Code

The model of an anomaly detector for HiLumi LHC magnets based on Recurrent Neural Networks and adaptive quantization

This paper focuses on an examination of an applicability of Recurrent Neural Network models for detecting anomalous behavior of the CERN superconducting magnets. In order to conduct the experiments, the authors designed and implemented an adaptive signal quantization algorithm and a custom GRU-based detector and developed a method for the detector parameters selection. Three different datasets were used for testing the detector. Two artificially generated datasets were used to assess the raw performance of the system whereas the 231 MB dataset composed of the signals acquired from HiLumi magnets was intended for real-life experiments and model training. Several different setups of the developed anomaly detection system were evaluated and compared with state-of-the-art OC-SVM reference model operating on the same data. The OC-SVM model was equipped with a rich set of feature extractors accounting for a range of the input signal properties. It was determined in the course of the experiments that the detector, along with its supporting design methodology, reaches F1 equal or very close to 1 for almost all test sets. Due to the profile of the data, the best_length setup of the detector turned out to perform the best among all five tested configuration schemes of the detection system. The quantization parameters have the biggest impact on the overall performance of the detector with the best values of input/output grid equal to 16 and 8, respectively. The proposed solution of the detection significantly outperformed OC-SVM-based detector in most of the cases, with much more stable performance across all the datasets.

arXiv:1712.01696

Avaliação do método dialético na quantização de imagens multiespectrais

The unsupervised classification has a very important role in the analysis of multispectral images, given its ability to assist the extraction of a priori knowledge of images. Algorithms like k-means and fuzzy c-means has long been used in this task. Computational Intelligence has proven to be an important field to assist in building classifiers optimized according to the quality of the grouping of classes and the evaluation of the quality of vector quantization. Several studies have shown that Philosophy, especially the Dialectical Method, has served as an important inspiration for the construction of new computational methods. This paper presents an evaluation of four methods based on the Dialectics: the Objective Dialectical Classifier and the Dialectical Optimization Method adapted to build a version of k-means with optimal quality indices; each of them is presented in two versions: a canonical version and another version obtained by applying the Principle of Maximum Entropy. These methods were compared to k-means, fuzzy c-means and Kohonen's self-organizing maps. The results showed that the methods based on Dialectics are robust to noise, and quantization can achieve results as good as those obtained with the Kohonen map, considered an optimal quantizer.

arXiv:1712.01695

Triagem virtual de imagens de imuno-histoquímica usando redes neurais artificiais e espectro de padrões

The importance of organizing medical images according to their nature, application and relevance is increasing. Furhermore, a previous selection of medical images can be useful to accelerate the task of analysis by pathologists. Herein this work we propose an image classifier to integrate a CBIR (Content-Based Image Retrieval) selection system. This classifier is based on pattern spectra and neural networks. Feature selection is performed using pattern spectra and principal component analysis, whilst image classification is based on multilayer perceptrons and a composition of self-organizing maps and learning vector quantization. These methods were applied for content selection of immunohistochemical images of placenta and newdeads lungs. Results demonstrated that this approach can reach reasonable classification performance.

arXiv:1712.01694

Fuzzy-Based Dialectical Non-Supervised Image Classification and Clustering

The materialist dialectical method is a philosophical investigative method to analyze aspects of reality. These aspects are viewed as complex processes composed by basic units named poles, which interact with each other. Dialectics has experienced considerable progress in the 19th century, with Hegel's dialectics and, in the 20th century, with the works of Marx, Engels, and Gramsci, in Philosophy and Economics. The movement of poles through their contradictions is viewed as a dynamic process with intertwined phases of evolution and revolutionary crisis. In order to build a computational process based on dialectics, the interaction between poles can be modeled using fuzzy membership functions. Based on this assumption, we introduce the Objective Dialectical Classifier (ODC), a non-supervised map for classification based on materialist dialectics and designed as an extension of fuzzy c-means classifier. As a case study, we used ODC to classify 181 magnetic resonance synthetic multispectral images composed by proton density, $T_1$- and $T_2$-weighted synthetic brain images. Comparing ODC to k-means, fuzzy c-means, and Kohonen's self-organized maps, concerning with image fidelity indexes as estimatives of quantization distortion, we proved that ODC can reach almost the same quantization performance as optimal non-supervised classifiers like Kohonen's self-organized maps.

arXiv:1712.01048

Adaptive Quantization for Deep Neural Network

In recent years Deep Neural Networks (DNNs) have been rapidly developed in various applications, together with increasingly complex architectures. The performance gain of these DNNs generally comes with high computational costs and large memory consumption, which may not be affordable for mobile platforms. Deep model quantization can be used for reducing the computation and memory costs of DNNs, and deploying complex DNNs on mobile equipment. In this work, we propose an optimization framework for deep model quantization. First, we propose a measurement to estimate the effect of parameter quantization errors in individual layers on the overall model prediction accuracy. Then, we propose an optimization process based on this measurement for finding optimal quantization bit-width for each layer. This is the first work that theoretically analyse the relationship between parameter quantization errors of individual layers and model accuracy. Our new quantization algorithm outperforms previous quantization optimization methods, and achieves 20-40% higher compression rate compared to equal bit-width quantization at the same model prediction accuracy.

arXiv:1712.00955

Composite Quantization

This paper studies the compact coding approach to approximate nearest neighbor search. We introduce a composite quantization framework. It uses the composition of several ($M$) elements, each of which is selected from a different dictionary, to accurately approximate a $D$-dimensional vector, thus yielding accurate search, and represents the data vector by a short code composed of the indices of the selected elements in the corresponding dictionaries. Our key contribution lies in introducing a near-orthogonality constraint, which makes the search efficiency is guaranteed as the cost of the distance computation is reduced to $O(M)$ from $O(D)$ through a distance table lookup scheme. The resulting approach is called near-orthogonal composite quantization. We theoretically justify the equivalence between near-orthogonal composite quantization and minimizing an upper bound of a function formed by jointly considering the quantization error and the search cost according to a generalized triangle inequality. We empirically show the efficacy of the proposed approach over several benchmark datasets. In addition, we demonstrate the superior performances in other three applications: combination with inverted multi-index, quantizing the query for mobile search, and inner-product similarity search.

arXiv:1712.02824

Stacked Denoising Autoencoders and Transfer Learning for Immunogold Particles Detection and Recognition

In this paper we present a system for the detection of immunogold particles and a Transfer Learning (TL) framework for the recognition of these immunogold particles. Immunogold particles are part of a high-magnification method for the selective localization of biological molecules at the subcellular level only visible through Electron Microscopy. The number of immunogold particles in the cell walls allows the assessment of the differences in their compositions providing a tool to analise the quality of different plants. For its quantization one requires a laborious manual labeling (or annotation) of images containing hundreds of particles. The system that is proposed in this paper can leverage significantly the burden of this manual task. For particle detection we use a LoG filter coupled with a SDA. In order to improve the recognition, we also study the applicability of TL settings for immunogold recognition. TL reuses the learning model of a source problem on other datasets (target problems) containing particles of different sizes. The proposed system was developed to solve a particular problem on maize cells, namely to determine the composition of cell wall ingrowths in endosperm transfer cells. This novel dataset as well as the code for reproducing our experiments is made publicly available. We determined that the LoG detector alone attained more than 84\% of accuracy with the F-measure. Developing immunogold recognition with TL also provided superior performance when compared with the baseline models augmenting the accuracy rates by 10\%.