arXiv subjects

Yao Shu

Publications and source records attributed to Yao Shu.

59 records · Page 4Linked to original sources

arXiv2020

Dynamic Routing Networks

The deployment of deep neural networks in real-world applications is mostly restricted by their high inference costs. Extensive efforts have been made to improve the accuracy with expert-designed or algorithm-searched architectures. However, the incremental improvement is typically achieved with increasingly more expensive models that only a small portion of input instances really need. Inference with a static architecture that processes all input instances via the same transformation would thus incur unnecessary computational costs. Therefore, customizing the model capacity in an instance-aware manner is much needed for higher inference efficiency. In this paper, we propose Dynamic Routing Networks (DRNets), which support efficient instance-aware inference by routing the input instance to only necessary transformation branches selected from a candidate set of branches for each connection between transformation nodes. The branch selection is dynamically determined via the corresponding branch importance weights, which are first generated from lightweight hypernetworks (RouterNets) and then recalibrated with Gumbel-Softmax before the selection. Extensive experiments show that DRNets can reduce a substantial amount of parameter size and FLOPs during inference with prediction performance comparable to state-of-the-art architectures.

cs.LG↗

arXiv2020

Effective and Efficient Dropout for Deep Convolutional Neural Networks

Convolutional Neural networks (CNNs) based applications have become ubiquitous, where proper regularization is greatly needed. To prevent large neural network models from overfitting, dropout has been widely used as an efficient regularization technique in practice. However, many recent works show that the standard dropout is ineffective or even detrimental to the training of CNNs. In this paper, we revisit this issue and examine various dropout variants in an attempt to improve existing dropout-based regularization techniques for CNNs. We attribute the failure of standard dropout to the conflict between the stochasticity of dropout and its following Batch Normalization (BN), and propose to reduce the conflict by placing dropout operations right before the convolutional operation instead of BN, or totally address this issue by replacing BN with Group Normalization (GN). We further introduce a structurally more suited dropout variant Drop-Conv2d, which provides more efficient and effective regularization for deep CNNs. These dropout variants can be readily integrated into the building blocks of CNNs and implemented in existing deep learning platforms. Extensive experiments on benchmark datasets including CIFAR, SVHN and ImageNet are conducted to compare the existing building blocks and the proposed ones with dropout training. Results show that our building blocks improve over state-of-the-art CNNs significantly, which is mainly due to the better regularization and implicit model ensemble effect.

cs.LG↗

arXiv2020

Understanding Architectures Learnt by Cell-based Neural Architecture Search

Neural architecture search (NAS) searches architectures automatically for given tasks, e.g., image classification and language modeling. Improving the search efficiency and effectiveness have attracted increasing attention in recent years. However, few efforts have been devoted to understanding the generated architectures. In this paper, we first reveal that existing NAS algorithms (e.g., DARTS, ENAS) tend to favor architectures with wide and shallow cell structures. These favorable architectures consistently achieve fast convergence and are consequently selected by NAS algorithms. Our empirical and theoretical study further confirms that their fast convergence derives from their smooth loss landscape and accurate gradient information. Nonetheless, these architectures may not necessarily lead to better generalization performance compared with other candidate architectures in the same search space, and therefore further improvement is possible by revising existing NAS algorithms.

cs.LG↗

arXiv2019

Efficient Memory Management for GPU-based Deep Learning Systems

GPU (graphics processing unit) has been used for many data-intensive applications. Among them, deep learning systems are one of the most important consumer systems for GPU nowadays. As deep learning applications impose deeper and larger models in order to achieve higher accuracy, memory management becomes an important research topic for deep learning systems, given that GPU has limited memory size. Many approaches have been proposed towards this issue, e.g., model compression and memory swapping. However, they either degrade the model accuracy or require a lot of manual intervention. In this paper, we propose two orthogonal approaches to reduce the memory cost from the system perspective. Our approaches are transparent to the models, and thus do not affect the model accuracy. They are achieved by exploiting the iterative nature of the training algorithm of deep learning to derive the lifetime and read/write order of all variables. With the lifetime semantics, we are able to implement a memory pool with minimal fragments. However, the optimization problem is NP-complete. We propose a heuristic algorithm that reduces up to 13.3% of memory compared with Nvidia's default memory pool with equal time complexity. With the read/write semantics, the variables that are not in use can be swapped out from GPU to CPU to reduce the memory footprint. We propose multiple swapping strategies to automatically decide which variable to swap and when to swap out (in), which reduces the memory cost by up to 34.2% without communication overhead.

cs.DC↗

arXiv2018

Randomness in Deconvolutional Networks for Visual Representation

Toward a deeper understanding on the inner work of deep neural networks, we investigate CNN (convolutional neural network) using DCN (deconvolutional network) and randomization technique, and gain new insights for the intrinsic property of this network architecture. For the random representations of an untrained CNN, we train the corresponding DCN to reconstruct the input images. Compared with the image inversion on pre-trained CNN, our training converges faster and the yielding network exhibits higher quality for image reconstruction. It indicates there is rich information encoded in the random features; the pre-trained CNN may discard information irrelevant for classification and encode relevant features in a way favorable for classification but harder for reconstruction. We further explore the property of the overall random CNN-DCN architecture. Surprisingly, images can be inverted with satisfactory quality. Extensive empirical evidence as well as theoretical analysis are provided.

cs.CV↗