With the success of deep neural networks, Neural Architecture Search (NAS) as a way of automatic model design has attracted wide attention. As training every child model from scratch is very time-consuming, recent works leverage weight-sharing to speed up the model evaluation procedure. These approaches greatly reduce computation by maintaining a single copy of weights on the super-net and share the weights among every child model. However, weight-sharing has no theoretical guarantee and its impact has not been well studied before. In this paper, we conduct comprehensive experiments to reveal the impact of weight-sharing: (1) The best-performing models from different runs or even from consecutive epochs within the same run have significant variance; (2) Even with high variance, we can extract valuable information from training the super-net with shared weights; (3) The interference between child models is a main factor that induces high variance; (4) Properly reducing the degree of weight sharing could effectively reduce variance and improve performance.
Weight sharing promises to make neural architecture search (NAS) tractable even on commodity hardware. Existing methods in this space rely on a diverse set of heuristics to design and train the shared-weight backbone network, a.k.a. the super-net. Since heuristics and hyperparameters substantially vary across different methods, a fair comparison between them can only be achieved by systematically analyzing the influence of these factors. In this paper, we therefore provide a systematic evaluation of the heuristics and hyperparameters that are frequently employed by weight-sharing NAS algorithm…
More transformer blocks with residual connections have recently achieved impressive results on various tasks. To achieve better performance with fewer trainable parameters, recent methods are proposed to go shallower by parameter sharing or model compressing along with the depth. However, weak modeling capacity limits their performance. Contrastively, going wider by inducing more trainable matrixes and parameters would produce a huge model requiring advanced parallelism to train and inference. In this paper, we propose a parameter-efficient framework, going wider instead of deeper. Specially, …
Gleamer（Gleamer公司）；Sorbonne Université（索邦大学）；CNRS（法国国家科学研究中心）；Institut des Systèmes Intelligents et de Robotique（智能系统与机器人研究所）
Weight-sharing (WS) has recently emerged as a paradigm to accelerate the automated search for efficient neural architectures, a process dubbed Neural Architecture Search (NAS). Although very appealing, this framework is not without drawbacks and several works have started to question its capabilities on small hand-crafted benchmarks. In this paper, we take advantage of the \nasbench dataset to challenge the efficiency of WS on a representative search space. By comparing a SOTA WS approach to a plain random search we show that, despite decent correlations between evaluations using weight-sharin…
Huawei（华为）；Samsung Research China, Beijing（三星中国研究院（北京））
Weight sharing based and predictor based methods are two major types of fast neural architecture search methods. In this paper, we propose to jointly use weight sharing and predictor in a unified framework. First, we construct a SuperNet in a weight-sharing way and probabilisticly sample architectures from the SuperNet. To increase the correctness of the evaluation of architectures, besides direct evaluation using the inherited weights, we further apply a few-shot predictor to assess the architecture on the other hand. The final evaluation of the architecture is the combination of direct evalu…
Efficient Neural Architecture Search methods based on weight sharing have shown good promise in democratizing Neural Architecture Search for computer vision models. There is, however, an ongoing debate whether these efficient methods are significantly better than random search. Here we perform a thorough comparison between efficient and random search methods on a family of progressively larger and more challenging search spaces for image classification and detection on ImageNet and COCO. While the efficacies of both methods are problem-dependent, our experiments demonstrate that there are larg…
Weight sharing, as an approach to speed up architecture performance estimation has received wide attention. Instead of training each architecture separately, weight sharing builds a supernet that assembles all the architectures as its submodels. However, there has been debate over whether the NAS process actually benefits from weight sharing, due to the gap between supernet optimization and the objective of NAS. To further understand the effect of weight sharing on NAS, we conduct a comprehensive analysis on five search spaces, including NAS-Bench-101, NAS-Bench-201, DARTS-CIFAR10, DARTS-PTB, …
Neural architecture search has attracted wide attentions in both academia and industry. To accelerate it, researchers proposed weight-sharing methods which first train a super-network to reuse computation among different operators, from which exponentially many sub-networks can be sampled and efficiently evaluated. These methods enjoy great advantages in terms of computational costs, but the sampled sub-networks are not guaranteed to be estimated precisely unless an individual training process is taken. This paper owes such inaccuracy to the inevitable mismatch between assembled network layers…
Huawei Inc.（华为公司）；Shanghai Jiao Tong University（上海交通大学）；Beijing University of Posts and Telecommunications（北京邮电大学）
Neural architecture search (NAS) has attracted increasing attentions in both academia and industry. In the early age, researchers mostly applied individual search methods which sample and evaluate the candidate architectures separately and thus incur heavy computational overheads. To alleviate the burden, weight-sharing methods were proposed in which exponentially many architectures share weights in the same super-network, and the costly training procedure is performed only once. These methods, though being much faster, often suffer the issue of instability. This paper provides a literature re…
One of the most critical problems in weight-sharing neural architecture search is the evaluation of candidate models within a predefined search space. In practice, a one-shot supernet is trained to serve as an evaluator. A faithful ranking certainly leads to more accurate searching results. However, current methods are prone to making misjudgments. In this paper, we prove that their biased evaluation is due to inherent unfairness in the supernet training. In view of this, we propose two levels of constraints: expectation fairness and strict fairness. Particularly, strict fairness ensures equal…
Scaled dot-product attention applies a softmax function on the scaled dot-product of queries and keys to calculate weights and then multiplies the weights and values. In this work, we study how to improve the learning of scaled dot-product attention to improve the accuracy of DETR. Our method is based on the following observations: using ground truth foreground-background mask (GT Fg-Bg Mask) as additional cues in the weights/values learning enables learning much better weights/values; with better weights/values, better values/weights can be learned. We propose a triple-attention module in whi…
To discover powerful yet compact models is an important goal of neural architecture search. Previous two-stage one-shot approaches are limited by search space with a fixed depth. It seems handy to include an additional skip connection in the search space to make depths variable. However, it creates a large range of perturbation during supernet training and it has difficulty giving a confident ranking for subnetworks. In this paper, we discover that skip connections bring about significant feature inconsistency compared with other operations, which potentially degrades the supernet performance.…
Weight sharing promises to make neural architecture search (NAS) tractable even on commodity hardware. Existing methods in this space rely on a diverse set of heuristics to design and train the shared-weight backbone network, a.k.a. the super-net. Since heuristics substantially vary across different methods and have not been carefully studied, it is unclear to which extent they impact super-net training and hence the weight-sharing NAS algorithms. In this paper, we disentangle super-net training from the search algorithm, isolate 14 frequently-used training heuristics, and evaluate them over t…
Michigan State University（密歇根州立大学）；AInnovation（旷视）
The use of automatic methods, often referred to as Neural Architecture Search (NAS), in designing neural network architectures has recently drawn considerable attention. In this work, we present an efficient NAS approach, named HM- NAS, that generalizes existing weight sharing based NAS approaches. Existing weight sharing based NAS approaches still adopt hand-designed heuristics to generate architecture candidates. As a consequence, the space of architecture candidates is constrained in a subset of all possible architectures, making the architecture search results sub-optimal. HM-NAS addresses…
Hong Kong University of Science and Technology（香港科技大学）；The University of Hong Kong（香港大学）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）
Neural Architecture Search (NAS) has shown great potentials in finding better neural network designs. Sample-based NAS is the most reliable approach which aims at exploring the search space and evaluating the most promising architectures. However, it is computationally very costly. As a remedy, the one-shot approach has emerged as a popular technique for accelerating NAS using weight-sharing. However, due to the weight-sharing of vastly different networks, the one-shot approach is less reliable than the sample-based approach. In this work, we propose BONAS (Bayesian Optimized Neural Architectu…
South China University of Technology（华南理工大学）；Tencent AI Lab（腾讯人工智能实验室）；University of Texas at Arlington（德克萨斯大学阿灵顿分校）
Neural architecture search (NAS) has gained increasing attention in the community of architecture design. One of the key factors behind the success lies in the training efficiency created by the weight sharing (WS) technique. However, WS-based NAS methods often suffer from a performance disturbance (PD) issue. That is, the training of subsequent architectures inevitably disturbs the performance of previously trained architectures due to the partially shared weights. This leads to inaccurate performance estimation for the previous architectures, which makes it hard to learn a good search strate…
In this paper, we propose the differentiable channel sparsity search (DCSS) for convolutional neural networks. Unlike traditional channel pruning algorithms which require users to manually set prune ratios for each convolutional layer, DCSS automatically searches the optimal combination of sparsities. Inspired by the differentiable architecture search (DARTS), we draw lessons from the continuous relaxation and leverage the gradient information to balance the computational cost and metrics. Since directly applying the scheme of DARTS causes shape mismatching and excessive memory consumption, we…
Computer Vision Laboratory, École Polytechnique Fédérale de Lausanne（洛桑联邦理工学院计算机视觉实验室）
The performance of a classifier trained on data coming from a specific domain typically degrades when applied to a related but different one. While annotating many samples from the new domain would address this issue, it is often too expensive or impractical. Domain Adaptation has therefore emerged as a solution to this problem; It leverages annotated data from a source domain, in which it is abundant, to train a classifier to operate in a target domain, in which it is either sparse or even lacking altogether. In this context, the recent trend consists of learning deep architectures whose weig…
Weight sharing has become a de facto standard in neural architecture search because it enables the search to be done on commodity hardware. However, recent works have empirically shown a ranking disorder between the performance of stand-alone architectures and that of the corresponding shared-weight networks. This violates the main assumption of weight-sharing NAS algorithms, thus limiting their effectiveness. We tackle this issue by proposing a regularization term that aims to maximize the correlation between the performance rankings of the shared-weight network and that of the standalone arc…
Facebook（脸书）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）
Weight-sharing neural architecture search (NAS) is an effective technique for automating efficient neural architecture design. Weight-sharing NAS builds a supernet that assembles all the architectures as its sub-networks and jointly trains the supernet with the sub-networks. The success of weight-sharing NAS heavily relies on distilling the knowledge of the supernet to the sub-networks. However, we find that the widely used distillation divergence, i.e., KL divergence, may lead to student sub-networks that over-estimate or under-estimate the uncertainty of the teacher supernet, leading to infe…
Convolutional neural networks (CNNs) are able to attain better visual recognition performance than fully connected neural networks despite having much fewer parameters due to their parameter sharing principle. Modern architectures usually contain a small number of fully-connected layers, often at the end, after multiple layers of convolutions. In some cases, most of the convolutions can be eliminated without suffering any loss in recognition performance. However, there is no solid recipe to detect the hidden subset of convolutional neurons that is responsible for the majority of the recognitio…
Modern computer vision is all about the possession of powerful image representations. Deeper and deeper convolutional neural networks have been built using larger and larger datasets and are made publicly available. A large swath of computer vision scientists use these pre-trained networks with varying degrees of successes in various tasks. Even though there is tremendous success in copying these networks, the representational space is not learnt from the target dataset in a traditional manner. One of the reasons for opting to use a pre-trained network over a network learnt from scratch is tha…
Hikvision Research Institute（海康威视研究院）；Zhejiang University（浙江大学）；University of Science and Technology of China（中国科学技术大学）
One-shot neural architecture search (NAS) applies weight-sharing supernet to reduce the unaffordable computation overhead of automated architecture designing. However, the weight-sharing technique worsens the ranking consistency of performance due to the interferences between different candidate networks. To address this issue, we propose a candidates enhancement method and progressive training pipeline to improve the ranking correlation of supernet. Specifically, we carefully redesign the sub-networks in the supernet and map the original supernet to a new one of high capacity. In addition, we…
University of Amsterdam（阿姆斯特丹大学）；Canadian Institute for Advanced Research (CIFAR)（加拿大高等研究院（CIFAR））
The success of deep learning in numerous application domains created the de- sire to run and train them on mobile devices. This however, conflicts with their computationally, memory and energy intense nature, leading to a growing interest in compression. Recent work by Han et al. (2015a) propose a pipeline that involves retraining, pruning and quantization of neural network weights, obtaining state-of-the-art compression rates. In this paper, we show that competitive compression rates can be achieved by using a version of soft weight-sharing (Nowlan & Hinton, 1992). Our method achieves both qu…
We present a framework for translating unlabeled images from one domain into analog images in another domain. We employ a progressively growing skip-connected encoder-generator structure and train it with a GAN loss for realistic output, a cycle consistency loss for maintaining same-domain translation identity, and a semantic consistency loss that encourages the network to keep the input semantic features in the output. We apply our framework on the task of translating face images, and show that it is capable of learning semantic mappings for face images with no supervised one-to-one image map…
In deep multi-task learning, weights of task-specific networks are shared between tasks to improve performance on each single one. Since the question, which weights to share between layers, is difficult to answer, human-designed architectures often share everything but a last task-specific layer. In many cases, this simplistic approach severely limits performance. Instead, we propose an algorithm to learn the assignment between a shared set of weights and task-specific layers. To optimize the non-differentiable assignment and at the same time train the differentiable weights, learning takes pl…
SenseTime Research（商汤科技研究院）；Beihang University（北京航空航天大学）；University of Sydney（悉尼大学）
The recent progress on automatically searching augmentation policies has boosted the performance substantially for various tasks. A key component of automatic augmentation search is the evaluation process for a particular augmentation policy, which is utilized to return reward and usually runs thousands of times. A plain evaluation process, which includes full model training and validation, would be time-consuming. To achieve efficiency, many choose to sacrifice evaluation reliability for speed. In this paper, we dive into the dynamics of augmented training of the model. This inspires us to de…
The University of Sydney（悉尼大学）；SenseTime Research（商汤科技研究院）；Tsinghua University（清华大学）
In one-shot weight sharing for NAS, the weights of each operation (at each layer) are supposed to be identical for all architectures (paths) in the supernet. However, this rules out the possibility of adjusting operation weights to cater for different paths, which limits the reliability of the evaluation results. In this paper, instead of counting on a single supernet, we introduce $K$-shot supernets and take their weights for each operation as a dictionary. The operation weight for each path is represented as a convex combination of items in a dictionary with a simplex code. This enables a ma…
We propose a simple and generic layer formulation that extends the properties of convolutional layers to any domain that can be described by a graph. Namely, we use the support of its adjacency matrix to design learnable weight sharing filters able to exploit the underlying structure of signals in the same fashion as for images. The proposed formulation makes it possible to learn the weights of the filter as well as a scheme that controls how they are shared across the graph. We perform validation experiments with image datasets and show that these filters offer performances comparable with co…
The current trend of pushing CNNs deeper with convolutions has created a pressing demand to achieve higher compression gains on CNNs where convolutions dominate the computation and parameter amount (e.g., GoogLeNet, ResNet and Wide ResNet). Further, the high energy consumption of convolutions limits its deployment on mobile devices. To this end, we proposed a simple yet effective scheme for compressing convolutions though applying k-means clustering on the weights, compression is achieved through weight-sharing, by only recording $K$ cluster centers and weight assignment indexes. We then intro…
Department of Computer Science, The University of Texas at Austin（德克萨斯大学奥斯汀分校计算机科学系）；Google（谷歌）
We propose a simple and efficient approach for training the BERT model. Our approach exploits the special structure of BERT that contains a stack of repeated modules (i.e., transformer encoders). Our proposed approach first trains BERT with the weights shared across all the repeated modules till some point. This is for learning the commonly shared component of weights across all repeated layers. We then stop weight sharing and continue training until convergence. We present theoretic insights for training by sharing weights then unsharing with analysis for simplified models. Empirical experime…
University of Louisiana at Lafayette（路易斯安那大学拉斐特分校）
In recent years, hypercomplex-inspired neural networks (HCNNs) have been used to improve deep learning architectures due to their ability to enable channel-based weight sharing, treat colors as a single entity, and improve representational coherence within the layers. The work described herein studies the effect of replacing existing layers in an Axial Attention network with their representationally coherent variants to assess the effect on image classification. We experiment with the stem of the network, the bottleneck layers, and the fully connected backend, by replacing them with representa…
The Hebrew University（希伯来大学）；Weizmann Institute of Science（魏茨曼科学研究学院）
Exploiting the great expressive power of Deep Neural Network architectures, relies on the ability to train them. While current theoretical work provides, mostly, results showing the hardness of this task, empirical evidence usually differs from this line, with success stories in abundance. A strong position among empirically successful architectures is captured by networks where extensive weight sharing is used, either by Convolutional or Recurrent layers. Additionally, characterizing specific aspects of different tasks, making them "harder" or "easier", is an interesting direction explored bo…
The University of Tokyo（东京大学）；National Institute of Advanced Industrial Science and Technology (AIST)（产业技术综合研究所）
Transformers have shown improved performance when compared to previous architectures for sequence processing such as RNNs. Despite their sizeable performance gains, as recently suggested, the model is computationally expensive to train and with a high parameter budget. In light of this, we explore parameter-sharing methods in Transformers with a specific focus on generative models. We perform an analysis of different parameter sharing/reduction methods and develop the Subformer. Our model combines sandwich-style parameter sharing, which overcomes naive cross-layer parameter sharing in generati…
Continued improvements in machine learning techniques offer exciting new opportunities through the use of larger models and larger training datasets. However, there is a growing need to offer these new capabilities on-board low-powered devices such as smartphones, wearables and other embedded environments where only low memory is available. Towards this, we consider methods to reduce the model size of Conformer-based speech recognition models which typically require models with greater than 100M parameters down to just $5$M parameters while minimizing impact on model quality. Such a model allo…
Chapman University（查普曼大学）；University of California, Irvine（加州大学欧文分校）
Weight-sharing is one of the pillars behind Convolutional Neural Networks and their successes. However, in physical neural systems such as the brain, weight-sharing is implausible. This discrepancy raises the fundamental question of whether weight-sharing is necessary. If so, to which degree of precision? If not, what are the alternatives? The goal of this study is to investigate these questions, primarily through simulations where the weight-sharing assumption is relaxed. Taking inspiration from neural circuitry, we explore the use of Free Convolutional Networks and neurons with variable conn…
Deep Residual Networks have reached the state of the art in many image processing tasks such image classification. However, the cost for a gain in accuracy in terms of depth and memory is prohibitive as it requires a higher number of residual blocks, up to double the initial value. To tackle this problem, we propose in this paper a way to reduce the redundant information of the networks. We share the weights of convolutional layers between residual blocks operating at the same spatial scale. The signal flows multiple times in the same convolutional layer. The resulting architecture, called Sha…
Ecole Centrale de Lyon（里昂中央理工学院）；Institut des Nanotechnologies de Lyon and CNRS（里昂纳米技术研究所及法国国家科学研究中心）
The computational workload involved in Convolutional Neural Networks (CNNs) is typically out of reach for low-power embedded devices. There are a large number of approximation techniques to address this problem. These methods have hyper-parameters that need to be optimized for each CNNs using design space exploration (DSE). The goal of this work is to demonstrate that the DSE phase time can easily explode for state of the art CNN. We thus propose the use of an optimized exploration process to drastically reduce the exploration time without sacrificing the quality of the output.
Convolutional Neural Networks (CNNs) are known to be significantly over-parametrized, and difficult to interpret, train and adapt. In this paper, we introduce a structural regularization across convolutional kernels in a CNN. In our approach, each convolution kernel is first decomposed as 2D dictionary atoms linearly combined by coefficients. The widely observed correlation and redundancy in a CNN hint a common low-rank structure among the decomposed coefficients, which is here further supported by our empirical observations. We then explicitly regularize CNN kernels by enforcing decomposed co…
Tuning hyperparameters is a crucial but arduous part of the machine learning pipeline. Hyperparameter optimization is even more challenging in federated learning, where models are learned over a distributed network of heterogeneous devices; here, the need to keep data on device and perform local training makes it difficult to efficiently train and evaluate configurations. In this work, we investigate the problem of federated hyperparameter tuning. We first identify key challenges and show how standard approaches may be adapted to form baselines for the federated setting. Then, by making a nove…
Georgia Institute of Technology（佐治亚理工学院）；Microsoft Azure AI（微软Azure人工智能部门）
Model ensemble is a popular approach to produce a low-variance and well-generalized model. However, it induces large memory and inference costs, which are often not affordable for real-world deployment. Existing work has resorted to sharing weights among models. However, when increasing the proportion of the shared weights, the resulting models tend to be similar, and the benefits of using model ensemble diminish. To retain ensemble benefits while maintaining a low memory cost, we propose a consistency-regularized ensemble learning approach based on perturbed models, named CAMERO. Specifically…
University of Cincinnati（辛辛那提大学）；University of Louisiana at Lafayette（路易斯安那大学拉法叶分校）
Long short-term memory (LSTM) is one of the robust recurrent neural network architectures for learning sequential data. However, it requires considerable computational power to learn and implement both software and hardware aspects. This paper proposed a novel LiteLSTM architecture based on reducing the LSTM computation components via the weights sharing concept to reduce the overall architecture computation cost and maintain the architecture performance. The proposed LiteLSTM can be significant for processing large data where time-consuming is crucial while hardware resources are limited, suc…
IBM Research（IBM研究院）；Stanford University（斯坦福大学）；University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）
The options framework is a popular approach for building temporally extended actions in reinforcement learning. In particular, the option-critic architecture provides general purpose policy gradient theorems for learning actions from scratch that are extended in time. However, past work makes the key assumption that each of the components of option-critic has independent parameters. In this work we note that while this key assumption of the policy gradient theorems of option-critic holds in the tabular case, it is always violated in practice for the deep function approximation setting. We thus…
Towards the Training of Deeper Predictive Coding Neural Networks
Institute of Logic and Computation, Vienna University of Technology（逻辑与计算研究所，维也纳技术大学）；University of Oxford（牛津大学）；VERSES AI Research Lab（VERSES AI研究实验室）
University of Vienna（维也纳大学）；Johann Radon Institute for Computational and Applied Mathematics, Austrian Academy of Sciences（奥地利科学院约翰·拉东计算与应用数学研究所）
Accurate numerical solutions for the Schrödinger equation are of utmost importance in quantum chemistry. However, the computational cost of current high-accuracy methods scales poorly with the number of interacting particles. Combining Monte Carlo methods with unsupervised training of neural networks has recently been proposed as a promising approach to overcome the curse of dimensionality in this setting and to obtain accurate wavefunctions for individual molecules at a moderately scaling computational cost. These methods currently do not exploit the regularity exhibited by wavefunctions with…
