Current methods for training recurrent neural networks are based on backpropagation through time, which requires storing a complete history of network states, and prohibits updating the weights `online' (after every timestep). Real Time Recurrent Learning (RTRL) eliminates the need for history storage and allows for online weight updates, but does so at the expense of computational costs that are quartic in the state size. This renders RTRL training intractable for all but the smallest networks, even ones that are made highly sparse.
We introduce the Sparse n-step Approximation (SnAp) to the RTRL influence matrix, which only keeps entries that are nonzero within n steps of the recurrent core. SnAp with n=1 is no more expensive than backpropagation, and we find that it substantially outperforms other RTRL approximations with comparable costs such as Unbiased Online Recurrent Optimization. For highly sparse networks, SnAp with n=2 remains tractable and can outperform backpropagation through time in terms of learning speed when updates are done online. SnAp becomes equivalent to RTRL when n is large.
University of Stuttgart（斯图加特大学）；ETAS GmbH（ETAS有限公司）
Modeling sequential data has become more and more important in practice. Some applications are autonomous driving, virtual sensors and weather forecasting. To model such systems, so called recurrent models are frequently used. In this paper we introduce several new Deep recurrent Gaussian process (DRGP) models based on the Sparse Spectrum Gaussian process (SSGP) and the improved version, called variational Sparse Spectrum Gaussian process (VSSGP). We follow the recurrent structure given by an existing DRGP based on a specific variational sparse Nyström approximation, the recurrent Gaussian pro…
Given a convolutional dictionary underlying a set of observed signals, can a carefully designed auto-encoder recover the dictionary in the presence of noise? We introduce an auto-encoder architecture, termed constrained recurrent sparse auto-encoder (CRsAE), that answers this question in the affirmative. Given an input signal and an approximate dictionary, the encoder finds a sparse approximation using FISTA. The decoder reconstructs the signal by applying the dictionary to the output of the encoder. The encoder and decoder in CRsAE parallel the sparse-coding and dictionary update steps in opt…
Recurrent neural networks (RNNs) are powerful and effective for processing sequential data. However, RNNs are usually considered "black box" models whose internal structure and learned parameters are not interpretable. In this paper, we propose an interpretable RNN based on the sequential iterative soft-thresholding algorithm (SISTA) for solving the sequential sparse recovery problem, which models a sequence of correlated observations with a sequence of sparse latent vectors. The architecture of the resulting SISTA-RNN is implicitly defined by the computational structure of SISTA, which result…
Recurrent Neural Networks (RNNs) are powerful tools for solving sequence-based problems, but their efficacy and execution time are dependent on the size of the network. Following recent work in simplifying these networks with model pruning and a novel mapping of work onto GPUs, we design an efficient implementation for sparse RNNs. We investigate several optimizations and tradeoffs: Lamport timestamps, wide memory loads, and a bank-aware weight layout. With these optimizations, we achieve speedups of over 6x over the next best algorithm for a hidden layer of size 2304, batch size of 4, and a d…
University of California, Berkeley（加州大学伯克利分校）；Microsoft Research（微软研究院）
Visual attention helps achieve robust perception under noise, corruption, and distribution shifts in human vision, which are areas where modern neural networks still fall short. We present VARS, Visual Attention from Recurrent Sparse reconstruction, a new attention formulation built on two prominent features of the human visual attention mechanism: recurrency and sparsity. Related features are grouped together via recurrent connections between neurons, with salient objects emerging via sparse regularization. VARS adopts an attractor network with recurrent connections that converges toward a st…
Courant Institute of Mathematical Sciences, New York University（纽约大学库朗数学科学研究所）
We present the discriminative recurrent sparse auto-encoder model, comprising a recurrent encoder of rectified linear units, unrolled for a fixed number of iterations, and connected to two linear decoders that reconstruct the input and predict its supervised classification. Training via backpropagation-through-time initially minimizes an unsupervised sparse reconstruction error; the loss function is then augmented with a discriminative term on the supervised classification. The depth implicit in the temporally-unrolled form allows the system to exhibit all the power of deep networks, while sub…
MILA, Université de Montréal（蒙特利尔大学 MILA）；Polytechnique Montréal（蒙特利尔理工学院）；HEC Montréal（蒙特利尔高等商学院）
A major drawback of backpropagation through time (BPTT) is the difficulty of learning long-term dependencies, coming from having to propagate credit information backwards through every single step of the forward computation. This makes BPTT both computationally impractical and biologically implausible. For this reason, full backpropagation through time is rarely used on long sequences, and truncated backpropagation through time is used as a heuristic. However, this usually leads to biased estimates of the gradient in which longer term dependencies are ignored. Addressing this issue, we propose…
Recurrent Neural Networks (RNNs) are used in state-of-the-art models in domains such as speech recognition, machine translation, and language modelling. Sparsity is a technique to reduce compute and memory requirements of deep learning models. Sparse RNNs are easier to deploy on devices and high-end server processors. Even though sparse operations need less compute and memory relative to their dense counterparts, the speed-up observed by using sparse operations is less than expected on different hardware platforms. In order to address this issue, we investigate two different approaches to indu…
University of Toronto（多伦多大学）；Vector Institute（向量研究所）
Recent advances in the sparse neural network literature have made it possible to prune many large feed forward and convolutional networks with only a small quantity of data. Yet, these same techniques often falter when applied to the problem of recovering sparse recurrent networks. These failures are quantitative: when pruned with recent techniques, RNNs typically obtain worse performance than they do under a simple random pruning scheme. The failures are also qualitative: the distribution of active weights in a pruned LSTM or GRU network tend to be concentrated in specific neurons and gates, …
Recurrent neural networks (RNNs) are omnipresent in sequence modeling tasks. Practical models usually consist of several layers of hundreds or thousands of neurons which are fully connected. This places a heavy computational and memory burden on hardware, restricting adoption in practical low-cost and low-power devices. Compared to fully convolutional models, the costly sequential operation of RNNs severely hinders performance on parallel hardware. This paper challenges the convention of full connectivity in RNNs. We study structurally sparse RNNs, showing that they are well suited for acceler…
We propose a new deep recurrent neural network (RNN) architecture for sequential signal reconstruction. Our network is designed by unfolding the iterations of the proximal gradient method that solves the l1-l1 minimization problem. As such, our network leverages by design that signals have a sparse representation and that the difference between consecutive signal representations is also sparse. We evaluate the proposed model in the task of reconstructing video frames from compressive measurements and show that it outperforms several state-of-the-art RNN models.
SparseProp: Efficient Event-Based Simulation and Training of Sparse Recurrent Spiking Neural Networks
Zuckerman Mind Brain Behavior Institute, Columbia University（哥伦比亚大学祖克曼心智脑行为研究所）；Max Planck Institute for Dynamics and Self-Organization（马克斯·普朗克动力学与自组织研究所）；Göttingen Campus Institute for Dynamics of Biological Networks（哥廷根校区生物网络动力学研究所）
Origin-destination (OD) matrices are often used in urban planning, where a city is partitioned into regions and an element (i, j) in an OD matrix records the cost (e.g., travel time, fuel consumption, or travel speed) from region i to region j. In this paper, we partition a day into multiple intervals, e.g., 96 15-min intervals and each interval is associated with an OD matrix which represents the costs in the interval; and we consider sparse and stochastic OD matrices, where the elements represent stochastic but not deterministic costs and some elements are missing due to lack of data between…
Conventional methods for human pose estimation either require a high degree of instrumentation, by relying on many inertial measurement units (IMUs), or constraint the recording space, by relying on extrinsic cameras. These deficits are tackled through the approach of human pose estimation from sparse IMU data. We define attention-oriented adjacency adaptive graph convolutional long-short term memory networks (A3GC-LSTM), to tackle human pose estimation based on six IMUs, through incorporating the human body graph structure directly into the network. The A3GC-LSTM combines both spatial and tem…
Institute for Neural Computation, Ruhr University Bochum（波鸿鲁尔大学神经计算研究所）；Royal Holloway, University of London（伦敦大学皇家霍洛威学院）；Faculty of Electrical and Computer Engineering, Technische Universität Dresden（德累斯顿工业大学电气与计算机工程学院）
Recurrent neural networks (RNNs) are well suited for solving sequence tasks in resource-constrained systems due to their expressivity and low computational requirements. However, there is still a need to bridge the gap between what RNNs are capable of in terms of efficiency and performance and real-world application requirements. The memory and computational requirements arising from propagating the activations of all the neurons at every time step to every connected neuron, together with the sequential dependence of activations, contribute to the inefficiency of training and using RNNs. We pr…
Korea Advanced Institute of Science and Technology (KAIST)（韩国科学技术院（KAIST））
Throughout this paper, we focus on the improvement of the direct feedback alignment (DFA) algorithm and extend the usage of the DFA to convolutional and recurrent neural networks (CNNs and RNNs). Even though the DFA algorithm is biologically plausible and has a potential of high-speed training, it has not been considered as the substitute for back-propagation (BP) due to the low accuracy in the CNN and RNN training. In this work, we propose a new DFA algorithm for BP-level accurate CNN and RNN training. Firstly, we divide the network into several modules and apply the DFA algorithm within the …
University Clinic Tübingen（图宾根大学医院）；International Max Planck Research School for Intelligent Systems（国际马克斯·普朗克智能系统研究学院）
Recurrent Neural Networks (RNNs) are popular models of brain function. The typical training strategy is to adjust their input-output behavior so that it matches that of the biological circuit of interest. Even though this strategy ensures that the biological and artificial networks perform the same computational task, it does not guarantee that their internal activity dynamics match. This suggests that the trained RNNs might end up performing the task employing a different internal computational mechanism, which would make them a suboptimal model of the biological circuit. In this work, we int…
Recurrent Neural Network (RNN) has been widely used to tackle a wide variety of language generation problems and are capable of attaining state-of-the-art (SOTA) performance. However despite its impressive results, the large number of parameters in the RNN model makes deployment to mobile and embedded devices infeasible. Driven by this problem, many works have proposed a number of pruning methods to reduce the sizes of the RNN model. In this work, we propose an end-to-end pruning method for image captioning models equipped with visual attention. Our proposed method is able to achieve sparsity …
Recurrent neural networks are powerful models for sequential data, able to represent complex dependencies in the sequence that simpler models such as hidden Markov models cannot handle. Yet they are notoriously hard to train. Here we introduce a training procedure using a gradient ascent in a Riemannian metric: this produces an algorithm independent from design choices such as the encoding of parameters and unit activities. This metric gradient ascent is designed to have an algorithmic cost close to backpropagation through time for sparsely connected networks. We use this procedure on gated le…
The objective of this work is to improve the accuracy of building demand forecasting. This is a more challenging task than grid level forecasting. For the said purpose, we develop a new technique called recurrent transform learning (RTL). Two versions are proposed. The first one (RTL) is unsupervised; this is used as a feature extraction tool that is further fed into a regression model. The second formulation embeds regression into the RTL framework leading to regressing recurrent transform learning (R2TL). Forecasting experiments have been carried out on three popular publicly available datas…
University of Electronic Science and Technology of China（电子科技大学）；The Chinese University of Hong Kong（香港中文大学）；Peng Cheng Lab（鹏城实验室）
Recurrent neural networks (RNNs) have recently achieved remarkable successes in a number of applications. However, the huge sizes and computational burden of these models make it difficult for their deployment on edge devices. A practically effective approach is to reduce the overall storage and computation costs of RNNs by network pruning techniques. Despite their successful applications, those pruning methods based on Lasso either produce irregular sparse patterns in weight matrices, which is not helpful in practical speedup. To address these issues, we propose structured pruning method thro…
Reservoir Computing (RC) is a well-known strategy for designing Recurrent Neural Networks featured by striking efficiency of training. The crucial aspect of RC is to properly instantiate the hidden recurrent layer that serves as dynamical memory to the system. In this respect, the common recipe is to create a pool of randomly and sparsely connected recurrent neurons. While the aspect of sparsity in the design of RC systems has been debated in the literature, it is nowadays understood mainly as a way to enhance the efficiency of computation, exploiting sparse matrix operations. In this paper, w…
Huawei Amsterdam Research Center（华为阿姆斯特丹研究中心）；National and Kapodistrian University of Athens（雅典国立卡波季斯特里安大学）
The ever increasing amount of GPS-equipped vehicles provides in real-time valuable traffic information for the roads traversed by the moving vehicles. In this way, a set of sparse and time evolving traffic reports is generated for each road. These time series are a valuable asset in order to forecast the future traffic condition. In this paper we present a deep learning framework that encodes the sparse recent traffic information and forecasts the future traffic condition. Our framework consists of a recurrent part and a decoder. The recurrent part employs an attention mechanism that encodes t…
University of Washington（华盛顿大学）；Applied Physics Laboratory, University of Washington（华盛顿大学应用物理实验室）
In this paper, we propose a novel recurrent neural network architecture for speech separation. This architecture is constructed by unfolding the iterations of a sequential iterative soft-thresholding algorithm (ISTA) that solves the optimization problem for sparse nonnegative matrix factorization (NMF) of spectrograms. We name this network architecture deep recurrent NMF (DR-NMF). The proposed DR-NMF network has three distinct advantages. First, DR-NMF provides better interpretability than other deep architectures, since the weights correspond to NMF model parameters, even after training. This…
National Research University Higher School of Economics（国立研究大学高等经济学院）；Kaspersky Lab（卡巴斯基实验室）；Lomonosov Moscow State University（莫斯科国立大学）；Yandex
Recurrent neural networks show state-of-the-art results in many text analysis tasks but often require a lot of memory to store their weights. Recently proposed Sparse Variational Dropout eliminates the majority of the weights in a feed-forward neural network without significant loss of quality. We apply this technique to sparsify recurrent neural networks. To account for recurrent specifics we also rely on Binary Variational Dropout for RNN. We report 99.5% sparsity level on sentiment analysis task without a quality drop and up to 87% sparsity level on language modeling task with slight loss o…
In sequence learning tasks such as language modelling, Recurrent Neural Networks must learn relationships between input features separated by time. State of the art models such as LSTM and Transformer are trained by backpropagation of losses into prior hidden states and inputs held in memory. This allows gradients to flow from present to past and effectively learn with perfect hindsight, but at a significant memory cost. In this paper we show that it is possible to train high performance recurrent networks using information that is local in time, and thereby achieve a significantly reduced mem…
HadamRNN: Binary and Sparse Ternary Orthogonal RNNs
Institut de Mathématiques de Toulouse（图卢兹数学研究所）；Institut de Recherche Technologique Saint Exupéry（圣埃克苏佩里技术研究所）；Université de Toulouse（图卢兹大学）；CNRS（法国国家科学研究中心）
While recurrent models have been effective in NLP tasks, their performance on context-free languages (CFLs) has been found to be quite weak. Given that CFLs are believed to capture important phenomena such as hierarchical structure in natural languages, this discrepancy in performance calls for an explanation. We study the performance of recurrent models on Dyck-n languages, a particularly important and well-studied class of CFLs. We find that while recurrent models generalize nearly perfectly if the lengths of the training and test strings are from the same range, they perform poorly if the t…
T-SHRED: Symbolic Regression for Regularization and Model Discovery with Transformer Shallow Recurrent Decoders
Electrical and Computer Engineering, University of Washington（华盛顿大学电气与计算机工程系）；Applied Mathematics, University of Washington（华盛顿大学应用数学系）；Computer Science & Engineering, University of Washington（华盛顿大学计算机科学与工程系）
Accelerating Linear Recurrent Neural Networks for the Edge with Unstructured Sparsity
Neuromorphic Computing Lab, Intel Corporation, USA；Institute of Informatics, LMU Munich, Germany；Bernoulli Institute \& CogniGron, University of Groningen, Netherlands
Surface Electromyography (sEMG/EMG) is to record muscles' electrical activity from a restricted area of the skin by using electrodes. The sEMG-based gesture recognition is extremely sensitive of inter-session and inter-subject variances. We propose a model and a deep-learning-based domain adaptation method to approximate the domain shift for recognition accuracy enhancement. Analysis performed on sparse and HighDensity (HD) sEMG public datasets validate that our approach outperforms state-of-the-art methods.
National Research University Higher School of Economics（国立研究型大学高等经济学院）；Samsung AI Center（三星人工智能中心）
Bayesian methods have been successfully applied to sparsify weights of neural networks and to remove structure units from the networks, e. g. neurons. We apply and further develop this approach for gated recurrent architectures. Specifically, in addition to sparsification of individual weights and neurons, we propose to sparsify preactivations of gates and information flow in LSTM. It makes some gates and information flow components constant, speeds up forward pass and improves compression. Moreover, the resulting structure of gate sparsity is interpretable and depends on the task. Code is ava…
The emergence of brain-inspired neuromorphic computing as a paradigm for edge AI is motivating the search for high-performance and efficient spiking neural networks to run on this hardware. However, compared to classical neural networks in deep learning, current spiking neural networks lack competitive performance in compelling areas. Here, for sequential and streaming tasks, we demonstrate how a novel type of adaptive spiking recurrent neural network (SRNN) is able to achieve state-of-the-art performance compared to other spiking neural networks and almost reach or exceed the performance of c…
Argonne National Laboratory（阿贡国家实验室）；University of Illinois at Chicago（伊利诺伊大学芝加哥分校）
Robot-assisted minimally invasive surgery is improving surgeon performance and patient outcomes. This innovation is also turning what has been a subjective practice into motion sequences that can be precisely measured. A growing number of studies have used machine learning to analyze video and kinematic data captured from surgical robots. In these studies, models are typically trained on benchmark datasets for representative surgical tasks to assess surgeon skill levels. While they have shown that novices and experts can be accurately classified, it is not clear whether machine learning can se…
Arizona State University（亚利桑那州立大学）；Shaanxi Normal University（陕西师范大学）
Reservoir computing systems, a class of recurrent neural networks, have recently been exploited for model-free, data-based prediction of the state evolution of a variety of chaotic dynamical systems. The prediction horizon demonstrated has been about half dozen Lyapunov time. Is it possible to significantly extend the prediction time beyond what has been achieved so far? We articulate a scheme incorporating time-dependent but sparse data inputs into reservoir computing and demonstrate that such rare "updates" of the actual state practically enable an arbitrarily long prediction horizon for a v…
CWI（CWI研究所）；Stichting IMEC Netherlands, Holst Centre（荷兰IMEC基金会荷尔斯特中心）；Univ of Amsterdam（阿姆斯特丹大学）；Rijksuniversiteit Groningen（格罗宁根大学）
Inspired by more detailed modeling of biological neurons, Spiking neural networks (SNNs) have been investigated both as more biologically plausible and potentially more powerful models of neural computation, and also with the aim of extracting biological neurons' energy efficiency; the performance of such networks however has remained lacking compared to classical artificial neural networks (ANNs). Here, we demonstrate how a novel surrogate gradient combined with recurrent networks of tunable and adaptive spiking neurons yields state-of-the-art for SNNs on challenging benchmarks in the time-do…
Sparsity in the structure of Neural Networks can lead to less energy consumption, less memory usage, faster computation times on convenient hardware, and automated machine learning. If sparsity gives rise to certain kinds of structure, it can explain automatically obtained features during learning. We provide insights into experiments in which we show how sparsity can be achieved through prior initialization, pruning, and during learning, and answer questions on the relationship between the structure of Neural Networks and their performance. This includes the first work of inducing priors from…
PySHRED: A Python package for SHallow REcurrent Decoding for sparse sensing, model reduction and scientific discovery
Department of Applied Computational Mathematical Sciences, University of Washington（应用计算数学科学系，华盛顿大学）；Department of Computer Science, University of Washington（计算机科学系，华盛顿大学）；Department of Mechanical Engineering, University of Washington（机械工程系，华盛顿大学）；Department of Energy, Nuclear Engineering Division, Politecnico di Milano（能源部，核工程系，米兰理工学院）；Department of Mechanical Engineering, Politecnico di Milano（机械工程系，米兰理工学院）；Department of Electrical and Computer Engineering, University of Washington（电气与计算机工程系，华盛顿大学）；Department of Applied Mathematics, University of Washington（应用数学系，华盛顿大学）
There are time series that are amenable to recurrent neural network (RNN) solutions when treated as sequences, but some series, e.g. asynchronous time series, provide a richer variation of feature types than current RNN cells take into account. In order to address such situations, we introduce a unified RNN that handles five different feature types, each in a different manner. Our RNN framework separates sequential features into two groups dependent on their frequency, which we call sparse and dense features, and which affect cell updates differently. Further, we also incorporate time features…
Arizona State University（亚利桑那州立大学）；Samsung Research America（三星美国研究院）
Recurrent neural networks (RNNs) achieve cutting-edge performance on a variety of problems. However, due to their high computational and memory demands, deploying RNNs on resource constrained mobile devices is a challenging task. To guarantee minimum accuracy loss with higher compression rate and driven by the mobile resource requirement, we introduce a novel model compression approach DirNet based on an optimized fast dictionary learning algorithm, which 1) dynamically mines the dictionary atoms of the projection dictionary matrix within layer to adjust the compression rate 2) adaptively chan…
National Physical Laboratory（国家物理实验室）；Mathematical Institute（数学研究所）；University of Oxford（牛津大学）；IIIT Delhi（德里印度理工学院）
We propose to use stochastic Riemannian coordinate descent on the orthogonal group for recurrent neural network training. The algorithm rotates successively two columns of the recurrent matrix, an operation that can be efficiently implemented as a multiplication by a Givens matrix. In the case when the coordinate is selected uniformly at random at each iteration, we prove the convergence of the proposed algorithm under standard assumptions on the loss function, stepsize and minibatch noise. In addition, we numerically demonstrate that the Riemannian gradient in recurrent neural network trainin…
