Researchers have proposed various activation functions. These activation functions help the deep network to learn non-linear behavior with a significant effect on training dynamics and task performance. The performance of these activations also depends on the initial state of the weight parameters, i.e., different initial state leads to a difference in the performance of a network. In this paper, we have proposed a cooperative initialization for training the deep network using ReLU activation function to improve the network performance. Our approach uses multiple activation functions in the initial few epochs for the update of all sets of weight parameters while training the network. These activation functions cooperate to overcome their drawbacks in the update of weight parameters, which in effect learn better "feature representation" and boost the network performance later. Cooperative initialization based training also helps in reducing the overfitting problem and does not increase the number of parameters, inference (test) time in the final model while improving the performance. Experiments show that our approach outperforms various baselines and, at the same time, performs well over various tasks such as classification and detection. The Top-1 classification accuracy of the model trained using our approach improves by 2.8% for VGG-16 and 2.1% for ResNet-56 on CIFAR-100 dataset.
A proper initialization of the weights in a neural network is critical to its convergence. Current insights into weight initialization come primarily from linear activation functions. In this paper, I develop a theory for weight initializations with non-linear activations. First, I derive a general weight initialization strategy for any neural network using activation functions differentiable at 0. Next, I derive the weight initialization strategy for the Rectified Linear Unit (RELU), and provide theoretical insights into why the Xavier initialization is a poor choice with RELU activations. My…
The interpretability of neural networks (NNs) is a challenging but essential topic for transparency in the decision-making process using machine learning. One of the reasons for the lack of interpretability is random weight initialization, where the input is randomly embedded into a different feature space in each layer. In this paper, we propose an interpretation method for a deep multilayer perceptron, which is the most general architecture of NNs, based on identity initialization (namely, initialization using identity matrices). The proposed method allows us to analyze the contribution of e…
The weight initialization and the activation function of deep neural networks have a crucial impact on the performance of the training procedure. An inappropriate selection can lead to the loss of information of the input during forward propagation and the exponential vanishing/exploding of gradients during back-propagation. Understanding the theoretical properties of untrained random networks is key to identifying which deep networks may be trained successfully as recently demonstrated by Schoenholz et al. (2017) who showed that for deep feedforward neural networks only a specific choice of h…
Department of EECS, Oregon State University（俄勒冈州立大学电子工程与计算机科学系）
Training a neural network (NN) depends on multiple factors, including but not limited to the initial weights. In this paper, we focus on initializing deep NN parameters such that it performs better, comparing to random or zero initialization. We do this by reducing the process of initialization into an SMT solver. Previous works consider certain activation functions on small NNs, however the studied NN is a deep network with different activation functions. Our experiments show that the proposed approach for parameter initialization achieves better performance comparing to randomly initialized …
University of Technology Sydney（悉尼科技大学）；Northwestern University（西北大学）
The prevailing thinking is that orthogonal weights are crucial to enforcing dynamical isometry and speeding up training. The increase in learning speed that results from orthogonal initialization in linear networks has been well-proven. However, while the same is believed to also hold for nonlinear networks when the dynamical isometry condition is satisfied, the training dynamics behind this contention have not been thoroughly explored. In this work, we study the dynamics of ultra-wide networks across a range of architectures, including Fully Connected Networks (FCNs) and Convolutional Neural …
Pennsylvania State University（宾夕法尼亚州立大学）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）
Weight initialization plays an important role in training neural networks and also affects tremendous deep learning applications. Various weight initialization strategies have already been developed for different activation functions with different neural networks. These initialization algorithms are based on minimizing the variance of the parameters between layers and might still fail when neural networks are deep, e.g., dying ReLU. To address this challenge, we study neural networks from a nonlinear computation point of view and propose a novel weight initialization strategy that is based on…
Intrinsic training dynamics of deep neural networks
ENS - PSL Univ.（巴黎-萨克雷大学）；CNRS, ENS - PSL Univ.（国家科学研究中心、巴黎-萨克雷大学）；Inria, CNRS, ENS de Lyon, Université Claude Bernard Lyon 1, LIP, UMR 5668（法国国家信息与自动化技术研究所、国家科学研究中心、里昂大学、 Claude Bernard 大学、LIP 实验室、UMR 5668）
Deep Linear Network Training Dynamics from Random Initialization: Data, Width, Depth, and Hyperparameter Transfer
John Paulson School of Engineering and Applied Sciences, Harvard University（哈佛大学约翰·保罗森工程与应用科学学校）；Center for Brain Sciences（脑科学研究中心）；Kempner Institute（凯普纳研究所）
Using an extended and formalized version of the Q/C map analysis of Poole et al. (2016), along with Neural Tangent Kernel theory, we identify the main pathologies present in deep networks that prevent them from training fast and generalizing to unseen data, and show how these can be avoided by carefully controlling the "shape" of the network's initialization-time kernel function. We then develop a method called Deep Kernel Shaping (DKS), which accomplishes this using a combination of precise parameter initialization, activation function transformations, and small architectural tweaks, all of w…
In this work, we have analyzed the problem of relative pose initialization between two satellites: a chaser and a non-cooperating target. The analysis has been targeted to two close-range methods based on a monocular camera system: the Sharma-Ventura-D'Amico (SVD) method and the silhouette matching method. Both methods are based on a priori knowledge of the target geometry, but neither fiducial markers nor a priori range measurements or state information are needed. The tests were carried out using a 2U CubeSat mock-up as target attached to a motorized rotary stage to simulate its relative mot…
Boston University（波士顿大学）；Sorbonne Université（索邦大学）；Ecole Polytechnique（巴黎综合理工学院）
Dedicated neural network (NN) architectures have been designed to handle specific data types (such as CNN for images or RNN for text), which ranks them among state-of-the-art methods for dealing with these data. Unfortunately, no architecture has been found for dealing with tabular data yet, for which tree ensemble methods (tree boosting, random forests) usually show the best predictive performances. In this work, we propose a new sparse initialization technique for (potentially deep) multilayer perceptrons (MLP): we first train a tree-based procedure to detect feature interactions and use the…
Numerous research efforts have been made to stabilize the training of the Generative Adversarial Networks (GANs), such as through regularization and architecture design. However, we identify the instability can also arise from the fragile balance at the early stage of adversarial learning. This paper proposes the CoopInit, a simple yet effective cooperative learning-based initialization strategy that can quickly learn a good starting point for GANs, with a very small computation overhead during training. The proposed algorithm consists of two learning stages: (i) Cooperative initialization sta…
With the proliferation of deep convolutional neural network (CNN) algorithms for mobile processing, limited precision quantization has become an essential tool for CNN efficiency. Consequently, various works have sought to design fixed precision quantization algorithms and quantization-focused optimization techniques that minimize quantization induced performance degradation. However, there is little concrete understanding of how various CNN design decisions/best practices affect quantized inference behaviour. Weight initialization strategies are often associated with solving issues such as va…
California Institute of Technology（加州理工学院）；Georgia Institute of Technology（佐治亚理工学院）；NVIDIA（英伟达）
Deep neural networks are usually initialized with random weights, with adequately selected initial variance to ensure stable signal propagation during training. However, selecting the appropriate variance becomes challenging especially as the number of layers grows. In this work, we replace random weight initialization with a fully deterministic initialization scheme, viz., ZerO, which initializes the weights of networks with only zeros and ones (up to a normalization factor), based on identity and Hadamard transforms. Through both theoretical and empirical studies, we demonstrate that ZerO is…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Microsoft（微软）
We study in this paper how to initialize the parameters of multinomial logistic regression (a fully connected layer followed with softmax and cross entropy loss), which is widely used in deep neural network (DNN) models for classification problems. As logistic regression is widely known not having a closed-form solution, it is usually randomly initialized, leading to several deficiencies especially in transfer learning where all the layers except for the last task-specific layer are initialized using a pre-trained model. The deficiencies include slow convergence speed, possibility of stuck in …
In this paper, we present a novel approach to perform deep neural networks layer-wise weight initialization using Linear Discriminant Analysis (LDA). Typically, the weights of a deep neural network are initialized with: random values, greedy layer-wise pre-training (usually as Deep Belief Network or as auto-encoder) or by re-using the layers from another network (transfer learning). Hence, many training epochs are needed before meaningful weights are learned, or a rather similar dataset is required for seeding a fine-tuning of transfer learning. In this paper, we describe how to turn an LDA in…
Biomediq A/S（Biomediq公司）；Cerebriu A/S（Cerebriu公司）；University of Copenhagen（哥本哈根大学）；University College London（伦敦大学学院）；King’s College London（伦敦国王学院）
Weight initialization is important for faster convergence and stability of deep neural networks training. In this paper, a robust initialization method is developed to address the training instability in long short-term memory (LSTM) networks. It is based on a normalized random initialization of the network weights that aims at preserving the variance of the network input and output in the same range. The method is applied to standard LSTMs for univariate time series regression and to LSTMs robust to missing values for multivariate disease progression modeling. The results show that in all cas…
Univ. Bourgogne Franche-Comté (UBFC)（勃艮第-弗朗什-孔泰大学（UBFC））；FEMTO-ST Institute（FEMTO-ST研究院）；American University of Beirut（贝鲁特美国大学）；University of Bejaia（贝贾亚大学）
Clustering is an unsupervised machine learning method grouping data samples into clusters of similar objects. In practice, clustering has been used in numerous applications such as banking customers profiling, document retrieval, image segmentation, and e-commerce recommendation engines. However, the existing clustering techniques present significant limitations, from which is the dependability of their stability on the initialization parameters (e.g. number of clusters, centroids). Different solutions were presented in the literature to overcome this limitation (i.e. internal and external val…
University of São Paulo（圣保罗大学）；Ghent University（根特大学）
The deep learning literature is continuously updated with new architectures and training techniques. However, weight initialization is overlooked by most recent research, despite some intriguing findings regarding random weights. On the other hand, recent works have been approaching Network Science to understand the structure and dynamics of Artificial Neural Networks (ANNs) after training. Therefore, in this work, we analyze the centrality of neurons in randomly initialized networks. We show that a higher neuronal strength variance may decrease performance, while a lower neuronal strength var…
Active visual tracking of space non-cooperative object is significant for future intelligent spacecraft to realise space debris removal, asteroid exploration, autonomous rendezvous and docking. However, existing works often consider this task into different subproblems (e.g. image preprocessing, feature extraction and matching, position and pose estimation, control law design) and optimize each module alone, which are trivial and sub-optimal. To this end, we propose an end-to-end active visual tracking method based on DQN algorithm, named as DRLAVT. It can guide the chasing spacecraft approach…
Beijing University of Posts and Telecommunications（北京邮电大学）
Activation function is crucial to the recent successes of deep neural networks. In this paper, we first propose a new activation function, Multiple Parametric Exponential Linear Units (MPELU), aiming to generalize and unify the rectified and exponential linear units. As the generalized form, MPELU shares the advantages of Parametric Rectified Linear Unit (PReLU) and Exponential Linear Unit (ELU), leading to better classification performance and convergence property. In addition, weight initialization is very important to train very deep networks. The existing methods laid a solid foundation fo…
Carnegie Mellon University（卡内基梅隆大学）；Faculdade de Engenharia da Universidade do Porto（波尔图大学工程学院）；INESC TEC（葡萄牙系统与计算机工程研究所）
Is there an initialization for deep networks that requires no learning? ExplainFix adopts two design principles: the "fixed filters" principle that all spatial filter weights of convolutional neural networks can be fixed at initialization and never learned, and the "nimbleness" principle that only few network parameters suffice. We contribute (a) visual model-based explanations, (b) speed and accuracy gains, and (c) novel tools for deep convolutional neural networks. ExplainFix gives key insights that spatially fixed networks should have a steered initialization, that spatial convolution layer…
Binary neural networks improve computationally efficiency of deep models with a large margin. However, there is still a performance gap between a successful full-precision training and binary training. We bring some insights about why this accuracy drop exists and call for a better understanding of binary network geometry. We start with analyzing full-precision neural networks with ReLU activation and compare it with its binarized version. This comparison suggests to initialize networks with random bias, a counter-intuitive remedy.
University of Central Florida（中佛罗里达大学）；Toyota Motor North America R&D InfoTech Labs（丰田北美研发信息技术实验室）
The recent advancements in communication and computational systems has led to significant improvement of situational awareness in connected and autonomous vehicles. Computationally efficient neural networks and high speed wireless vehicular networks have been some of the main contributors to this improvement. However, scalability and reliability issues caused by inherent limitations of sensory and communication systems are still challenging problems. In this paper, we aim to mitigate the effects of these limitations by introducing the concept of feature sharing for cooperative object detection…
BITS-Pilani Hyderabad（比拉理工学院海得拉巴校区）；Indian Institute of Science（印度科学理工学院）
Next generation deep neural networks for classification hosted on embedded platforms will rely on fast, efficient, and accurate learning algorithms. Initialization of weights in learning networks has a great impact on the classification accuracy. In this paper we focus on deriving good initial weights by modeling the error function of a deep neural network as a high-dimensional landscape. We observe that due to the inherent complexity in its algebraic structure, such an error function may conform to general results of the statistics of large systems. To this end we apply some results from Rand…
University of Wollongong（卧龙岗大学）；Motovis Inc（魔视智能科技有限公司）；Chinese Academy of Sciences（中国科学院）；National University of Defense Technology（国防科技大学）
A novel deep neural network training paradigm that exploits the conjoint information in multiple heterogeneous sources is proposed. Specifically, in a RGB-D based action recognition task, it cooperatively trains a single convolutional neural network (named c-ConvNet) on both RGB visual features and depth features, and deeply aggregates the two kinds of features for action recognition. Differently from the conventional ConvNet that learns the deep separable features for homogeneous modality-based classification with only one softmax loss function, the c-ConvNet enhances the discriminative power…
University of California at Riverside（加州大学河滨分校）；Toyota North America R&D Labs（丰田北美研发实验室）
3D object detection plays a fundamental role in enabling autonomous driving, which is regarded as the significant key to unlocking the bottleneck of contemporary transportation systems from the perspectives of safety, mobility, and sustainability. Most of the state-of-the-art (SOTA) object detection methods from point clouds are developed based on a single onboard LiDAR, whose performance will be inevitably limited by the range and occlusion, especially in dense traffic scenarios. In this paper, we propose \textit{PillarGrid}, a novel cooperative perception method fusing information from multi…
Batch normalization (BN) is comprised of a normalization component followed by an affine transformation and has become essential for training deep neural networks. Standard initialization of each BN in a network sets the affine transformation scale and shift to 1 and 0, respectively. However, after training we have observed that these parameters do not alter much from their initialization. Furthermore, we have noticed that the normalization process can still yield overly large values, which is undesirable for training. We revisit the BN formulation and present a new initialization method and u…
Impact of Initialization on Intra-subject Pediatric Brain MR Image Registration: A Comparative Analysis between SyN ANTs and Deep Learning-Based Approaches
Polytechnique Montréal（蒙特利尔理工学院）；Ste-Justine Hospital University Centre（圣贾斯汀医院大学中心）；Université de Strasbourg（斯特拉斯堡大学）；CNRS（法国国家科学研究中心）
