We consider the problem of distributed mean estimation (DME), in which $n$ machines are each given a local $d$-dimensional vector $x_v \in \mathbb{R}^d$, and must cooperate to estimate the mean of their inputs $\mu = \frac 1n\sum_{v = 1}^n x_v$, while minimizing total communication cost. DME is a fundamental construct in distributed machine learning, and there has been considerable work on variants of this problem, especially in the context of distributed variance reduction for stochastic gradients in parallel SGD. Previous work typically assumes an upper bound on the norm of the input vectors, and achieves an error bound in terms of this norm. However, in many real applications, the input vectors are concentrated around the correct output $\mu$, but $\mu$ itself has large norm. In such cases, previous output error bounds perform poorly. In this paper, we show that output error bounds need not depend on input norm. We provide a method of quantization which allows distributed mean estimation to be performed with solution quality dependent only on the distance between inputs, not on input norm, and show an analogous result for distributed variance reduction. The technique is based on a new connection with lattice theory. We also provide lower bounds showing that the communication to error trade-off of our algorithms is asymptotically optimal. As the lattices achieving optimal bounds under $\ell_2$-norm can be computationally impractical, we also present an extension which leverages easy-to-use cubic lattices, and is loose only up to a logarithmic factor in $d$. We show experimentally that our method yields practical improvements for common applications, relative to prior approaches.
FPGAs provide a flexible and efficient platform to accelerate rapidly-changing algorithms for computer vision. The majority of existing work focuses on accelerating image classification, while other fundamental vision problems, including object detection and instance segmentation, have not been adequately addressed. Compared with image classification, detection problems are more sensitive to the spatial variance of objects, and therefore, require specialized convolutions to aggregate spatial information. To address this, recent work proposes dynamic deformable convolution to augment regular convolutions. Regular convolutions process a fixed grid of pixels across all the spatial locations in an image, while dynamic deformable convolutions may access arbitrary pixels in the image and the access pattern is input-dependent and varies per spatial location. These properties lead to inefficient memory accesses of inputs with existing hardware. In this work, we first investigate the overhead of the deformable convolution on embedded FPGA SoCs, and then show the accuracy-latency tradeoffs for a set of algorithm modifications including full versus depthwise, fixed-shape, and limited-range. These modifications benefit the energy efficiency for embedded devices in general as they reduce the compute complexity. We then build an efficient object detection network with modified deformable convolutions and quantize the network using state-of-the-art quantization methods. We implement a unified hardware engine on FPGA to support all the operations in the network. Preliminary experiments show that little accuracy is compromised and speedup can be achieved with our co-design optimization for the deformable convolution.
In this work we present a new framework for neural networks compression with fine-tuning, which we called Neural Network Compression Framework (NNCF). It leverages recent advances of various network compression methods and implements some of them, such as sparsity, quantization, and binarization. These methods allow getting more hardware-friendly models which can be efficiently run on general-purpose hardware computation units (CPU, GPU) or special Deep Learning accelerators. We show that the developed methods can be successfully applied to a wide range of models to accelerate the inference time while keeping the original accuracy. The framework can be used within the training samples, which are supplied with it, or as a standalone package that can be seamlessly integrated into the existing training code with minimal adaptations. Currently, a PyTorch version of NNCF is available as a part of OpenVINO Training Extensions at https://github.com/openvinotoolkit/nncf.
Recommender system models often represent various sparse features like users, items, and categorical features via embeddings. A standard approach is to map each unique feature value to an embedding vector. The size of the produced embedding table grows linearly with the size of the vocabulary. Therefore, a large vocabulary inevitably leads to a gigantic embedding table, creating two severe problems: (i) making model serving intractable in resource-constrained environments; (ii) causing overfitting problems. In this paper, we seek to learn highly compact embeddings for large-vocab sparse features in recommender systems (recsys). First, we show that the novel Differentiable Product Quantization (DPQ) approach can generalize to recsys problems. In addition, to better handle the power-law data distribution commonly seen in recsys, we propose a Multi-Granular Quantized Embeddings (MGQE) technique which learns more compact embeddings for infrequent items. We seek to provide a new angle to improve recommendation performance with compact model sizes. Extensive experiments on three recommendation tasks and two datasets show that we can achieve on par or better performance, with only ~20% of the original model size.
We propose a novel algorithm for quantizing continuous latent representations in trained models. Our approach applies to deep probabilistic models, such as variational autoencoders (VAEs), and enables both data and model compression. Unlike current end-to-end neural compression methods that cater the model to a fixed quantization scheme, our algorithm separates model design and training from quantization. Consequently, our algorithm enables "plug-and-play" compression with variable rate-distortion trade-off, using a single trained model. Our algorithm can be seen as a novel extension of arithmetic coding to the continuous domain, and uses adaptive quantization accuracy based on estimates of posterior uncertainty. Our experimental results demonstrate the importance of taking into account posterior uncertainties, and show that image compression with the proposed algorithm outperforms JPEG over a wide range of bit rates using only a single standard VAE. Further experiments on Bayesian neural word embeddings demonstrate the versatility of the proposed method.
Deep neural networks (DNNs) have been proven to outperform classical methods on several machine learning benchmarks. However, they have high computational complexity and require powerful processing units. Especially when deployed on embedded systems, model size and inference time must be significantly reduced. We propose SYMOG (symmetric mixture of Gaussian modes), which significantly decreases the complexity of DNNs through low-bit fixed-point quantization. SYMOG is a novel soft quantization method such that the learning task and the quantization are solved simultaneously. During training the weight distribution changes from an unimodal Gaussian distribution to a symmetric mixture of Gaussians, where each mean value belongs to a particular fixed-point mode. We evaluate our approach with different architectures (LeNet5, VGG7, VGG11, DenseNet) on common benchmark data sets (MNIST, CIFAR-10, CIFAR-100) and we compare with state-of-the-art quantization approaches. We achieve excellent results and outperform 2-bit state-of-the-art performance with an error rate of only 5.71% on CIFAR-10 and 27.65% on CIFAR-100.
Despite progress in training neural networks for lossy image compression, current approaches fail to maintain both perceptual quality and abstract features at very low bitrates. Encouraged by recent success in learning discrete representations with Vector Quantized Variational Autoencoders (VQ-VAEs), we motivate the use of a hierarchy of VQ-VAEs to attain high factors of compression. We show that the combination of stochastic quantization and hierarchical latent structure aids likelihood-based image compression. This leads us to introduce a novel objective for training hierarchical VQ-VAEs. Our resulting scheme produces a Markovian series of latent variables that reconstruct images of high-perceptual quality which retain semantically meaningful features. We provide qualitative and quantitative evaluations on the CelebA and MNIST datasets.
Neural network quantization methods often involve simulating the quantization process during training, making the trained model highly dependent on the target bit-width and precise way quantization is performed. Robust quantization offers an alternative approach with improved tolerance to different classes of data-types and quantization policies. It opens up new exciting applications where the quantization process is not static and can vary to meet different circumstances and implementations. To address this issue, we propose a method that provides intrinsic robustness to the model against a broad range of quantization processes. Our method is motivated by theoretical arguments and enables us to store a single generic model capable of operating at various bit-widths and quantization policies. We validate our method's effectiveness on different ImageNet models.
We analyze the effect of quantizing weights and activations of neural networks on their loss and derive a simple regularization scheme that improves robustness against post-training quantization. By training quantization-ready networks, our approach enables storing a single set of weights that can be quantized on-demand to different bit-widths as energy and memory requirements of the application change. Unlike quantization-aware training using the straight-through estimator that only targets a specific bit-width and requires access to training data and pipeline, our regularization-based method paves the way for "on the fly'' post-training quantization to various bit-widths. We show that by modeling quantization as a $\ell_\infty$-bounded perturbation, the first-order term in the loss expansion can be regularized using the $\ell_1$-norm of gradients. We experimentally validate the effectiveness of our regularization scheme on different architectures on CIFAR-10 and ImageNet datasets.
We propose precision gating (PG), an end-to-end trainable dynamic dual-precision quantization technique for deep neural networks. PG computes most features in a low precision and only a small proportion of important features in a higher precision to preserve accuracy. The proposed approach is applicable to a variety of DNN architectures and significantly reduces the computational cost of DNN execution with almost no accuracy loss. Our experiments indicate that PG achieves excellent results on CNNs, including statically compressed mobile-friendly networks such as ShuffleNet. Compared to the state-of-the-art prediction-based quantization schemes, PG achieves the same or higher accuracy with 2.4$\times$ less compute on ImageNet. PG furthermore applies to RNNs. Compared to 8-bit uniform quantization, PG obtains a 1.2% improvement in perplexity per word with 2.7$\times$ computational cost reduction on LSTM on the Penn Tree Bank dataset. Code is available at: https://github.com/cornell-zhang/dnn-gating
Backbone architectures of most binary networks are well-known floating point architectures such as the ResNet family. Questioning that the architectures designed for floating point networks would not be the best for binary networks, we propose to search architectures for binary networks (BNAS) by defining a new search space for binary architectures and a novel search objective. Specifically, based on the cell based search method, we define the new search space of binary layer types, design a new cell template, and rediscover the utility of and propose to use the Zeroise layer instead of using it as a placeholder. The novel search objective diversifies early search to learn better performing binary architectures. We show that our proposed method searches architectures with stable training curves despite the quantization error inherent in binary networks. Quantitative analyses demonstrate that our searched architectures outperform the architectures used in state-of-the-art binary networks and outperform or perform on par with state-of-the-art binary networks that employ various techniques other than architectural changes.
This research aims at identifying the unknown emotion using speaker cues. In this study, we identify the unknown emotion using a two-stage framework. The first stage focuses on identifying the speaker who uttered the unknown emotion, while the next stage focuses on identifying the unknown emotion uttered by the recognized speaker in the prior stage. This proposed framework has been evaluated on an Arabic Emirati-accented speech database uttered by fifteen speakers per gender. Mel-Frequency Cepstral Coefficients (MFCCs) have been used as the extracted features and Hidden Markov Model (HMM) has been utilized as the classifier in this work. Our findings demonstrate that emotion recognition accuracy based on the two-stage framework is greater than that based on the one-stage approach and the state-of-the-art classifiers and models such as Gaussian Mixture Model (GMM), Support Vector Machine (SVM), and Vector Quantization (VQ). The average emotion recognition accuracy based on the two-stage approach is 67.5%, while the accuracy reaches to 61.4%, 63.3%, 64.5%, and 61.5%, based on the one-stage approach, GMM, SVM, and VQ, respectively. The achieved results based on the two-stage framework are very close to those attained in subjective assessment by human listeners.
Recent neural text-to-speech (TTS) models with fine-grained latent features enable precise control of the prosody of synthesized speech. Such models typically incorporate a fine-grained variational autoencoder (VAE) structure, extracting latent features at each input token (e.g., phonemes). However, generating samples with the standard VAE prior often results in unnatural and discontinuous speech, with dramatic prosodic variation between tokens. This paper proposes a sequential prior in a discrete latent space which can generate more naturally sounding samples. This is accomplished by discretizing the latent features using vector quantization (VQ), and separately training an autoregressive (AR) prior model over the result. We evaluate the approach using listening tests, objective metrics of automatic speech recognition (ASR) performance, and measurements of prosody attributes. Experimental results show that the proposed model significantly improves the naturalness in random sample generation. Furthermore, initial experiments demonstrate that randomly sampling from the proposed model can be used as data augmentation to improve the ASR performance.
We envision that in the near future, humanoid robots would share home space and assist us in our daily and routine activities through object manipulations. One of the fundamental technologies that need to be developed for robots is to enable them to detect objects and recognize them for effective manipulations and take real-time decisions involving those objects. In this paper, we describe a novel architecture that enables multiple low-compute NAO robots to perform real-time detection, recognition and localization of objects in its camera view and take programmable actions based on the detected objects. The proposed algorithm for object detection and localization is an empirical modification of YOLOv3, based on indoor experiments in multiple scenarios, with a smaller weight size and lesser computational requirements. Quantization of the weights and re-adjusting filter sizes and layer arrangements for convolutions improved the inference time for low-resolution images from the robot s camera feed. YOLOv3 was chosen after a comparative study of bounding box algorithms was performed with an objective to choose one that strikes the perfect balance among information retention, low inference time and high accuracy for real-time object detection and localization. The architecture also comprises of an effective end-to-end pipeline to feed the real-time frames from the camera feed to the neural net and use its results for guiding the robot with customizable actions corresponding to the detected class labels.
Neural networks have demonstrably achieved state-of-the art accuracy using low-bitlength integer quantization, yielding both execution time and energy benefits on existing hardware designs that support short bitlengths. However, the question of finding the minimum bitlength for a desired accuracy remains open. We introduce a training method for minimizing inference bitlength at any granularity while maintaining accuracy. Namely, we propose a regularizer that penalizes large bitlength representations throughout the architecture and show how it can be modified to minimize other quantifiable criteria, such as number of operations or memory footprint. We demonstrate that our method learns thrifty representations while maintaining accuracy. With ImageNet, the method produces an average per layer bitlength of 4.13, 3.76 and 4.36 bits on AlexNet, ResNet18 and MobileNet V2 respectively, remaining within 2.0%, 0.5% and 0.5% of the base TOP-1 accuracy.
Instantaneous and on demand accuracy-efficiency trade-off has been recently explored in the context of neural networks slimming. In this paper, we propose a flexible quantization strategy, termed Switchable Precision neural Networks (SP-Nets), to train a shared network capable of operating at multiple quantization levels. At runtime, the network can adjust its precision on the fly according to instant memory, latency, power consumption and accuracy demands. For example, by constraining the network weights to 1-bit with switchable precision activations, our shared network spans from BinaryConnect to Binarized Neural Network, allowing to perform dot-products using only summations or bit operations. In addition, a self-distillation scheme is proposed to increase the performance of the quantized switches. We tested our approach with three different quantizers and demonstrate the performance of SP-Nets against independently trained quantized models in classification accuracy for Tiny ImageNet and ImageNet datasets using ResNet-18 and MobileNet architectures.
Over the last few years, Deep Neural Networks (DNNs) have become ubiquitous owing to their high accuracy on real-world tasks. However, this increase in accuracy comes at the cost of computationally expensive models leading to higher prediction latencies. Prior efforts to reduce this latency such as quantization, model distillation, and any-time prediction models typically trade-off accuracy for performance. In this work, we observe that caching intermediate layer outputs can help us avoid running all the layers of a DNN for a sizeable fraction of inference requests. We find that this can potentially reduce the number of effective layers by half for 91.58% of CIFAR-10 requests run on ResNet-18. We present Freeze Inference, a system that introduces approximate caching at each intermediate layer and we discuss techniques to reduce the cache size and improve the cache hit rate. Finally, we discuss some of the open research challenges in realizing such a design.
Consider the following distributed optimization scenario. A worker has access to training data that it uses to compute the gradients while a server decides when to stop iterative computation based on its target accuracy or delay constraints. The server receives all its information about the problem instance from the worker via a rate-limited noiseless communication channel. We introduce the principle we call Differential Quantization (DQ) that prescribes compensating the past quantization errors to direct the descent trajectory of a quantized algorithm towards that of its unquantized counterpart. Assuming that the objective function is smooth and strongly convex, we prove that Differentially Quantized Gradient Descent (DQ-GD) attains a linear contraction factor of $\max\{\sigma_{\mathrm{GD}}, \rho_n 2^{-R}\}$, where $\sigma_{\mathrm{GD}}$ is the contraction factor of unquantized gradient descent (GD), $\rho_n \geq 1$ is the covering efficiency of the quantizer, and $R$ is the bitrate per problem dimension $n$. Thus at any $R\geq\log_2 \rho_n /\sigma_{\mathrm{GD}}$ bits, the contraction factor of DQ-GD is the same as that of unquantized GD, i.e., there is no loss due to quantization. We show that no algorithm within a certain class can converge faster than $\max\{\sigma_{\mathrm{GD}}, 2^{-R}\}$. Since quantizers exist with $\rho_n \to 1$ as $n \to \infty$ (Rogers, 1963), this means that DQ-GD is asymptotically optimal. The principle of differential quantization continues to apply to gradient methods with momentum such as Nesterov's accelerated gradient descent, and Polyak's heavy ball method. For these algorithms as well, if the rate is above a certain threshold, there is no loss in contraction factor obtained by the differentially quantized algorithm compared to its unquantized counterpart. Experimental results on least-squares problems validate our theoretical analysis.
Despite the huge success of Deep Neural Networks (DNNs) in a wide spectrum of machine learning and data mining tasks, recent research shows that this powerful tool is susceptible to maliciously crafted adversarial examples. Up until now, adversarial training has been the most successful defense against adversarial attacks. To increase adversarial robustness, a DNN can be trained with a combination of benign and adversarial examples generated by first-order methods. However, in state-of-the-art first-order attacks, adversarial examples with sign gradients retain the sign information of each gradient component but discard the relative magnitude between components. In this work, we replace sign gradients with quantized gradients. Gradient quantization not only preserves the sign information, but also keeps the relative magnitude between components. Experiments show white-box first-order attacks with quantized gradients outperform their variants with sign gradients on multiple datasets. Notably, our BLOB\_QG attack achieves an accuracy of $88.32\%$ on the secret MNIST model from the MNIST Challenge and it outperforms all other methods on the leaderboard of white-box attacks.
Image hash algorithms generate compact binary representations that can be quickly matched by Hamming distance, thus become an efficient solution for large-scale image retrieval. This paper proposes RV-SSDH, a deep image hash algorithm that incorporates the classical VLAD (vector of locally aggregated descriptors) architecture into neural networks. Specifically, a novel neural network component is formed by coupling a random VLAD layer with a latent hash layer through a transform layer. This component can be combined with convolutional layers to realize a hash algorithm. We implement RV-SSDH as a point-wise algorithm that can be efficiently trained by minimizing classification error and quantization loss. Comprehensive experiments show this new architecture significantly outperforms baselines such as NetVLAD and SSDH, and offers a cost-effective trade-off in the state-of-the-art. In addition, the proposed random VLAD layer leads to satisfactory accuracy with low complexity, thus shows promising potentials as an alternative to NetVLAD.