We study the neural network (NN) compression problem, viewing the tension between the compression ratio and NN performance through the lens of rate-distortion theory. We choose a distortion metric that reflects the effect of NN compression on the model output and derive the tradeoff between rate (compression) and distortion. In addition to characterizing theoretical limits of NN compression, this formulation shows that \emph{pruning}, implicitly or explicitly, must be a part of a good compression algorithm. This observation bridges a gap between parts of the literature pertaining to NN and data compression, respectively, providing insight into the empirical success of model pruning. Finally, we propose a novel pruning strategy derived from our information-theoretic formulation and show that it outperforms the relevant baselines on CIFAR-10 and ImageNet datasets.
Unleashing the full potential of massive MIMO in FDD mode by reducing the overhead of CSI feedback has recently garnered attention. Numerous deep learning for massive MIMO CSI feedback approaches have demonstrated their efficiency and potential. However, most existing methods improve accuracy at the cost of computational complexity and the accuracy decreases significantly as the CSI compression rate increases. This paper presents a novel neural network CLNet tailored for CSI feedback problem based on the intrinsic properties of CSI. CLNet proposes a forge complex-valued input layer to process signals and utilizes attention mechanism to enhance the performance of the network. The experiment result shows that CLNet outperforms the state-of-the-art method by average accuracy improvement of 5.41\% in both outdoor and indoor scenarios with average 24.1\% less computational overhead. Codes for deep learning-based CSI feedback CLNet are available at GitHub.
Principal Component Analysis (PCA) is well known for its capability of dimension reduction and data compression. However, when using PCA for compressing/reconstructing images, images need to be recast to vectors. The vectorization of images makes some correlation constraints of neighboring pixels and spatial information lost. To deal with the drawbacks of the vectorizations adopted by PCA, we used small neighborhoods of each pixel to form compounded pixels and use a tensorial version of PCA, called TPCA (Tensorial Principal Component Analysis), to compress and reconstruct a compounded image of compounded pixels. Our experiments on public data show that TPCA compares favorably with PCA in compressing and reconstructing images. We also show in our experiments that the performance of TPCA increases when the order of compounded pixels increases.
Petabytes of data are generated each day by emerging Internet of Things (IoT), but only few of them can be finally collected and used for Machine Learning (ML) purposes due to the apprehension of data & privacy leakage, which seriously retarding ML's growth. To alleviate this problem, Federated learning is proposed to perform model training by multiple clients' combined data without the dataset sharing within the cluster. Nevertheless, federated learning introduces massive communication overhead as the synchronized data in each epoch is of the same size as the model, and thereby leading to a low communication efficiency. Consequently, variant methods mainly focusing on the communication rounds reduction and data compression are proposed to reduce the communication overhead of federated learning. In this paper, we propose Overlap-FedAvg, a framework that parallels the model training phase with model uploading & downloading phase, so that the latter phase can be totally covered by the former phase. Compared to vanilla FedAvg, Overlap-FedAvg is further developed with a hierarchical computing strategy, a data compensation mechanism and a nesterov accelerated gradients~(NAG) algorithm. Besides, Overlap-FedAvg is orthogonal to many other compression methods so that they can be applied together to maximize the utilization of the cluster. Furthermore, the theoretical analysis is provided to prove the convergence of the proposed Overlap-FedAvg framework. Extensive experiments on both conventional and recurrent tasks with multiple models and datasets also demonstrate that the proposed Overlap-FedAvg framework substantially boosts the federated learning process.
Density estimation, compression and data generation are crucial tasks in artificial intelligence. Variational Auto-Encoders (VAEs) constitute a single framework to achieve these goals. Here, we present a novel class of generative models, called self-supervised Variational Auto-Encoder (selfVAE), that utilizes deterministic and discrete variational posteriors. This class of models allows to perform both conditional and unconditional sampling, while simplifying the objective function. First, we use a single self-supervised transformation as a latent variable, where a transformation is either downscaling or edge detection. Next, we consider a hierarchical architecture, i.e., multiple transformations, and we show its benefits compared to the VAE. The flexibility of selfVAE in data reconstruction finds a particularly interesting use case in data compression tasks, where we can trade-off memory for better data quality, and vice-versa. We present performance of our approach on three benchmark image data (Cifar10, Imagenette64, and CelebA).
BIRCH clustering is a widely known approach for clustering, that has influenced much subsequent research and commercial products. The key contribution of BIRCH is the Clustering Feature tree (CF-Tree), which is a compressed representation of the input data. As new data arrives, the tree is eventually rebuilt to increase the compression. Afterward, the leaves of the tree are used for clustering. Because of the data compression, this method is very scalable. The idea has been adopted for example for k-means, data stream, and density-based clustering. Clustering features used by BIRCH are simple summary statistics that can easily be updated with new data: the number of points, the linear sums, and the sum of squared values. Unfortunately, how the sum of squares is then used in BIRCH is prone to catastrophic cancellation. We introduce a replacement cluster feature that does not have this numeric problem, that is not much more expensive to maintain, and which makes many computations simpler and hence more efficient. These cluster features can also easily be used in other work derived from BIRCH, such as algorithms for streaming data. In the experiments, we demonstrate the numerical problem and compare the performance of the original algorithm compared to the improved cluster features.
Previous studies have confirmed the effectiveness of incorporating visual information into speech enhancement (SE) systems. Despite improved denoising performance, two problems may be encountered when implementing an audio-visual SE (AVSE) system: (1) additional processing costs are incurred to incorporate visual input and (2) the use of face or lip images may cause privacy problems. In this study, we propose a Lite AVSE (LAVSE) system to address these problems. The system includes two visual data compression techniques and removes the visual feature extraction network from the training model, yielding better online computation efficiency. Our experimental results indicate that the proposed LAVSE system can provide notably better performance than an audio-only SE system with a similar number of model parameters. In addition, the experimental results confirm the effectiveness of the two techniques for visual data compression.
Wavelets are well known for data compression, yet have rarely been applied to the compression of neural networks. This paper shows how the fast wavelet transform can be used to compress linear layers in neural networks. Linear layers still occupy a significant portion of the parameters in recurrent neural networks (RNNs). Through our method, we can learn both the wavelet bases and corresponding coefficients to efficiently represent the linear layers of RNNs. Our wavelet compressed RNNs have significantly fewer parameters yet still perform competitively with the state-of-the-art on synthetic and real-world RNN benchmarks. Wavelet optimization adds basis flexibility, without large numbers of extra weights. Source code is available at https://github.com/v0lta/Wavelet-network-compression.
With the arrival of several face-swapping applications such as FaceApp, SnapChat, MixBooth, FaceBlender and many more, the authenticity of digital media content is hanging on a very loose thread. On social media platforms, videos are widely circulated often at a high compression factor. In this work, we analyze several deep learning approaches in the context of deepfakes classification in high compression scenario and demonstrate that a proposed approach based on metric learning can be very effective in performing such a classification. Using less number of frames per video to assess its realism, the metric learning approach using a triplet network architecture proves to be fruitful. It learns to enhance the feature space distance between the cluster of real and fake videos embedding vectors. We validated our approaches on two datasets to analyze the behavior in different environments. We achieved a state-of-the-art AUC score of 99.2% on the Celeb-DF dataset and accuracy of 90.71% on a highly compressed Neural Texture dataset. Our approach is especially helpful on social media platforms where data compression is inevitable.
Vector quantization (VQ) techniques are widely used in similarity search for data compression, fast metric computation and etc. Originally designed for Euclidean distance, existing VQ techniques (e.g., PQ, AQ) explicitly or implicitly minimize the quantization error. In this paper, we present a new angle to analyze the quantization error, which decomposes the quantization error into norm error and direction error. We show that quantization errors in norm have much higher influence on inner products than quantization errors in direction, and small quantization error does not necessarily lead to good performance in maximum inner product search (MIPS). Based on this observation, we propose norm-explicit quantization (NEQ) --- a general paradigm that improves existing VQ techniques for MIPS. NEQ quantizes the norms of items in a dataset explicitly to reduce errors in norm, which is crucial for MIPS. For the direction vectors, NEQ can simply reuse an existing VQ technique to quantize them without modification. We conducted extensive experiments on a variety of datasets and parameter configurations. The experimental results show that NEQ improves the performance of various VQ techniques for MIPS, including PQ, OPQ, RQ and AQ.
Time series data compression is emerging as an important problem with the growth in IoT devices and sensors. Due to the presence of noise in these datasets, lossy compression can often provide significant compression gains without impacting the performance of downstream applications. In this work, we propose an error-bounded lossy compressor, LFZip, for multivariate floating-point time series data that provides guaranteed reconstruction up to user-specified maximum absolute error. The compressor is based on the prediction-quantization-entropy coder framework and benefits from improved prediction using linear models and neural networks. We evaluate the compressor on several time series datasets where it outperforms the existing state-of-the-art error-bounded lossy compressors. The code and data are available at https://github.com/shubhamchandak94/LFZip
The goal of lossy data compression is to reduce the storage cost of a data set $X$ while retaining as much information as possible about something ($Y$) that you care about. For example, what aspects of an image $X$ contain the most information about whether it depicts a cat? Mathematically, this corresponds to finding a mapping $X\to Z\equiv f(X)$ that maximizes the mutual information $I(Z,Y)$ while the entropy $H(Z)$ is kept below some fixed threshold. We present a method for mapping out the Pareto frontier for classification tasks, reflecting the tradeoff between retained entropy and class information. We first show how a random variable $X$ (an image, say) drawn from a class $Y\in\{1,...,n\}$ can be distilled into a vector $W=f(X)\in \mathbb{R}^{n-1}$ losslessly, so that $I(W,Y)=I(X,Y)$; for example, for a binary classification task of cats and dogs, each image $X$ is mapped into a single real number $W$ retaining all information that helps distinguish cats from dogs. For the $n=2$ case of binary classification, we then show how $W$ can be further compressed into a discrete variable $Z=g_\beta(W)\in\{1,...,m_\beta\}$ by binning $W$ into $m_\beta$ bins, in such a way that varying the parameter $\beta$ sweeps out the full Pareto frontier, solving a generalization of the Discrete Information Bottleneck (DIB) problem. We argue that the most interesting points on this frontier are "corners" maximizing $I(Z,Y)$ for a fixed number of bins $m=2,3...$ which can be conveniently be found without multiobjective optimization. We apply this method to the CIFAR-10, MNIST and Fashion-MNIST datasets, illustrating how it can be interpreted as an information-theoretically optimal image clustering algorithm.
Sequential data is being generated at an unprecedented pace in various forms, including text and genomic data. This creates the need for efficient compression mechanisms to enable better storage, transmission and processing of such data. To solve this problem, many of the existing compressors attempt to learn models for the data and perform prediction-based compression. Since neural networks are known as universal function approximators with the capability to learn arbitrarily complex mappings, and in practice show excellent performance in prediction tasks, we explore and devise methods to compress sequential data using neural network predictors. We combine recurrent neural network predictors with an arithmetic coder and losslessly compress a variety of synthetic, text and genomic datasets. The proposed compressor outperforms Gzip on the real datasets and achieves near-optimal compression for the synthetic datasets. The results also help understand why and where neural networks are good alternatives for traditional finite context models
Matrix factorization (MF) discovers latent features from observations, which has shown great promises in the fields of collaborative filtering, data compression, feature extraction, word embedding, etc. While many problem-specific optimization techniques have been proposed, alternating least square (ALS) remains popular due to its general applicability e.g. easy to handle positive-unlabeled inputs, fast convergence and parallelization capability. Current MF implementations are either optimized for a single machine or with a need of a large computer cluster but still are insufficient. This is because a single machine provides limited compute power for large-scale data while multiple machines suffer from the network communication bottleneck. To address the aforementioned challenge, accelerating ALS on graphics processing units (GPUs) is a promising direction. We propose the novel approach in enhancing the MF efficiency via both memory optimization and approximate computing. The former exploits GPU memory hierarchy to increase data reuse, while the later reduces unnecessary computing without hurting the convergence of learning algorithms. Extensive experiments on large-scale datasets show that our solution not only outperforms the competing CPU solutions by a large margin but also has a 2x-4x performance gain compared to the state-of-the-art GPU solutions. Our implementations are open-sourced and publicly available.
We describe the multi-GPU gradient boosting algorithm implemented in the XGBoost library (https://github.com/dmlc/xgboost). Our algorithm allows fast, scalable training on multi-GPU systems with all of the features of the XGBoost library. We employ data compression techniques to minimise the usage of scarce GPU memory while still allowing highly efficient implementation. Using our algorithm we show that it is possible to process 115 million training instances in under three minutes on a publicly available cloud computing instance. The algorithm is implemented using end-to-end GPU parallelism, with prediction, gradient calculation, feature quantisation, decision tree construction and evaluation phases all computed on device.
Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems.
When performing localization and mapping, working at the level of structure can be advantageous in terms of robustness to environmental changes and differences in illumination. This paper presents SegMap: a map representation solution to the localization and mapping problem based on the extraction of segments in 3D point clouds. In addition to facilitating the computationally intensive task of processing 3D point clouds, working at the level of segments addresses the data compression requirements of real-time single- and multi-robot systems. While current methods extract descriptors for the single task of localization, SegMap leverages a data-driven descriptor in order to extract meaningful features that can also be used for reconstructing a dense 3D map of the environment and for extracting semantic information. This is particularly interesting for navigation tasks and for providing visual feedback to end-users such as robot operators, for example in search and rescue scenarios. These capabilities are demonstrated in multiple urban driving and search and rescue experiments. Our method leads to an increase of area under the ROC curve of 28.3% over current state of the art using eigenvalue based features. We also obtain very similar reconstruction capabilities to a model specifically trained for this task. The SegMap implementation will be made available open-source along with easy to run demonstrations at www.github.com/ethz-asl/segmap. A video demonstration is available at https://youtu.be/CMk4w4eRobg.