arXiv subjects

Song Han

Publications and source records attributed to Song Han.

247 records · Page 14Linked to original sources

arXiv2017

Classification of Neurological Gait Disorders Using Multi-task Feature Learning

As our population ages, neurological impairments and degeneration of the musculoskeletal system yield gait abnormalities, which can significantly reduce quality of life. Gait rehabilitative therapy has been widely adopted to help patients maximize community participation and living independence. To further improve the precision and efficiency of rehabilitative therapy, more objective methods need to be developed based on sensory data. In this paper, an algorithmic framework is proposed to provide classification of gait disorders caused by two common neurological diseases, stroke and Parkinson's Disease (PD), from ground contact force (GCF) data. An advanced machine learning method, multi-task feature learning (MTFL), is used to jointly train classification models of a subject's gait in three classes, post-stroke, PD and healthy gait. Gait parameters related to mobility, balance, strength and rhythm are used as features for the classification. Out of all the features used, the MTFL models capture the more important ones per disease, which will help provide better objective assessment and therapy progress tracking. To evaluate the proposed methodology we use data from a human participant study, which includes five PD patients, three post-stroke patients, and three healthy subjects. Despite the diversity of abnormalities, the evaluation shows that the proposed approach can successfully distinguish post-stroke and PD gait from healthy gait, as well as post-stroke from PD gait, with Area Under the Curve (AUC) score of at least 0.96. Moreover, the methodology helps select important gait features to better understand the key characteristics that distinguish abnormal gaits and design personalized treatment.

cs.CV↗

arXiv2016

Analog electromagnetically induced transparency for circularly polarized wave using three dimensional chiral metamaterials

In this paper, we theoretically and experimentally demonstrate a three dimensional metamaterial that can motivate electromagnetic induced transparency (EIT) by using circular polarized wave as stimulations. The unit cell consists of a pair of metallic strips printed on both sides of the printed circuit board (PCB), where a conductive cylinder junction is used to connect the metal strips by drilling a hole inside the substrate. When a right circularly polarized wave is incident, destructive interference is excited between meta-atoms of the 3D structure, the transmission spectrum demonstrates a sharp transparency window. A coupled oscillator model and an electrical equivalent circuit model are applied to quantitatively and qualitatively analyze the coupling mechanism in the EIT-like metamaterial. Analysis in detail shows the EIT window's amplitude and frequency are modulated by changing the degree of symmetry breaking. The proposed metamaterial may achieve potential applications in developing chiral slow light devices.

physics.class-ph↗

arXiv2016

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

Recent research on deep neural networks has focused primarily on improving accuracy. For a given accuracy level, it is typically possible to identify multiple DNN architectures that achieve that accuracy level. With equivalent accuracy, smaller DNN architectures offer at least three advantages: (1) Smaller DNNs require less communication across servers during distributed training. (2) Smaller DNNs require less bandwidth to export a new model from the cloud to an autonomous car. (3) Smaller DNNs are more feasible to deploy on FPGAs and other hardware with limited memory. To provide all of these advantages, we propose a small DNN architecture called SqueezeNet. SqueezeNet achieves AlexNet-level accuracy on ImageNet with 50x fewer parameters. Additionally, with model compression techniques we are able to compress SqueezeNet to less than 0.5MB (510x smaller than AlexNet). The SqueezeNet architecture is available for download here: https://github.com/DeepScale/SqueezeNet

cs.CV↗

arXiv2016

EIE: Efficient Inference Engine on Compressed Deep Neural Network

State-of-the-art deep neural networks (DNNs) have hundreds of millions of connections and are both computationally and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources and power budgets. While custom hardware helps the computation, fetching weights from DRAM is two orders of magnitude more expensive than ALU operations, and dominates the required power. Previously proposed 'Deep Compression' makes it possible to fit large DNNs (AlexNet and VGGNet) fully in on-chip SRAM. This compression is achieved by pruning the redundant connections and having multiple connections share the same weight. We propose an energy efficient inference engine (EIE) that performs inference on this compressed network model and accelerates the resulting sparse matrix-vector multiplication with weight sharing. Going from DRAM to SRAM gives EIE 120x energy saving; Exploiting sparsity saves 10x; Weight sharing gives 8x; Skipping zero activations from ReLU saves another 3x. Evaluated on nine DNN benchmarks, EIE is 189x and 13x faster when compared to CPU and GPU implementations of the same DNN without compression. EIE has a processing power of 102GOPS/s working directly on a compressed network, corresponding to 3TOPS/s on an uncompressed network, and processes FC layers of AlexNet at 1.88x10^4 frames/sec with a power dissipation of only 600mW. It is 24,000x and 3,400x more energy efficient than a CPU and GPU respectively. Compared with DaDianNao, EIE has 2.9x, 19x and 3x better throughput, energy efficiency and area efficiency.

cs.CV↗

arXiv2016

Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources. To address this limitation, we introduce "deep compression", a three stage pipeline: pruning, trained quantization and Huffman coding, that work together to reduce the storage requirement of neural networks by 35x to 49x without affecting their accuracy. Our method first prunes the network by learning only the important connections. Next, we quantize the weights to enforce weight sharing, finally, we apply Huffman coding. After the first two steps we retrain the network to fine tune the remaining connections and the quantized centroids. Pruning, reduces the number of connections by 9x to 13x; Quantization then reduces the number of bits that represent each connection from 32 to 5. On the ImageNet dataset, our method reduced the storage required by AlexNet by 35x, from 240MB to 6.9MB, without loss of accuracy. Our method reduced the size of VGG-16 by 49x from 552MB to 11.3MB, again with no loss of accuracy. This allows fitting the model into on-chip SRAM cache rather than off-chip DRAM memory. Our compression method also facilitates the use of complex neural networks in mobile applications where application size and download bandwidth are constrained. Benchmarked on CPU, GPU and mobile GPU, compressed network has 3x to 4x layerwise speedup and 3x to 7x better energy efficiency.

cs.CV↗

arXiv2016

Generate Image Descriptions based on Deep RNN and Memory Cells for Images Features

Generating natural language descriptions for images is a challenging task. The traditional way is to use the convolutional neural network (CNN) to extract image features, followed by recurrent neural network (RNN) to generate sentences. In this paper, we present a new model that added memory cells to gate the feeding of image features to the deep neural network. The intuition is enabling our model to memorize how much information from images should be fed at each stage of the RNN. Experiments on Flickr8K and Flickr30K datasets showed that our model outperforms other state-of-the-art models with higher BLEU scores.

cs.CV↗

arXiv2015

Learning both Weights and Connections for Efficient Neural Networks

Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems. Also, conventional networks fix the architecture before training starts; as a result, training cannot improve the architecture. To address these limitations, we describe a method to reduce the storage and computation required by neural networks by an order of magnitude without affecting their accuracy by learning only the important connections. Our method prunes redundant connections using a three-step method. First, we train the network to learn which connections are important. Next, we prune the unimportant connections. Finally, we retrain the network to fine tune the weights of the remaining connections. On the ImageNet dataset, our method reduced the number of parameters of AlexNet by a factor of 9x, from 61 million to 6.7 million, without incurring accuracy loss. Similar experiments with VGG-16 found that the number of parameters can be reduced by 13x, from 138 million to 10.3 million, again with no loss of accuracy.

cs.NE↗

arXiv2013

A Security Protocol for the Identification and Data Encrypt Key Management of Secure Mobile Devices

In this paper, we proposed an identification and data encrypt key manage protocol that can be used in some security system based on such secure devices as secure USB memories or RFIDs, which are widely used for identifying persons or other objects recently. In general, the default functions of the security system using a mobile device are the authentication for the owner of the device and secure storage of data stored on the device. We proposed a security model that consists of the server and mobile devices in order to realize these security features. In this model we defined the secure communication protocol for the authentication and management of data encryption keys using a private key encryption algorithm with the public key between the server and mobile devices. In addition, we was performed the analysis for the attack to the communication protocol between the mobile device and server. Using the communication protocol, the system will attempt to authenticate the mobile device. The data decrypt key is transmitted only if the authentication process is successful. The data in the mobile device can be decrypted using the key. Our analysis proved that this Protocol ensures anonymity, prevents replay attacks and realizes the interactive identification between the security devices and the authentication server.

cs.CR↗

arXiv2013

Software Cognitive Information Measure based on Relation Between Structures

Cognitive complexity measures quantify human difficulty in understanding the source code based on cognitive informatics foundation. The discipline derives cognitive complexity on a basis of fundamental software factors i.e, inputs, outputs, and internal processing architecture. An approach to integrating Granular Computing into the new measure called Structured Cognitive Information Measure or SCIM. The proposed measure unifies and re-organizes complexity factors analogous to human cognitive process. However, according to the methodology of software and the scope of the variables, Information Complexity Number(ICN) of variables is depended on change of variable value and cognitive complexity is measured in several ways. In this paper, we define the Scope Information Complexity Number (SICN) and present the cognitive complexity based on functional decomposition of software, including theoretical validation through nine Weyuker's properties.

cs.SE↗

arXiv2012

Towards Approximate Model Checking DC and PDC Specifications

DC has proved to be a promising tool for the specification and verification of functional requirements on the design of hard real-time systems. Many works were devoted to develop effective techniques for checking the models of hard real-time systems against DC specifications. DC model checking theory is still evolving and yet there is no available tools supporting practical verifications due to the high undecidability of calculus and the great complexity of model checking. Present situation of PDC model checking is much worse than the one of DC model checking. In view of the results so far achieved, it is desirable to develop approximate model checking techniques for DC and PDC specifications. This work was motivated to develop approximate techniques checking automata models of hard real-time systems for DC and PDC specifications. Unlike previous works which only deal with decidable formulas, we want to develop approximate techniques covering whole DC and PDC formulas. The first results of our work, namely, approximate techniques checking real-time automata models of systems for LDI and PLDI specifications, are described in this paper.

cs.LO↗

arXiv2012

Robust Face Recognition using Local Illumination Normalization and Discriminant Feature Point Selection

Face recognition systems must be robust to the variation of various factors such as facial expression, illumination, head pose and aging. Especially, the robustness against illumination variation is one of the most important problems to be solved for the practical use of face recognition systems. Gabor wavelet is widely used in face detection and recognition because it gives the possibility to simulate the function of human visual system. In this paper, we propose a method for extracting Gabor wavelet features which is stable under the variation of local illumination and show experiment results demonstrating its effectiveness.

cs.LG↗

arXiv2012

A Survey on Low-cost RFID Authentication Protocols

This paper presents a survey on several RFID authentication protocols under low cost restrictions. Low cost RFID are mainly addressed with limited security and privacy protections. In this study, we explore several protocols with various authentication mechanisms found in literature that satisfy low cost restrictions. Assessments of these protocols are based on data protection, tracking protection, forward security. Finally, it is concluded that no single low cost RFID protocol fully meets the requirement of the given assessments. While a protocol satisfies one or two assessments, it fails to fully meet the requirement of the third assessment. This study provides a new insight in RFID literature which can be used particularly by small and medium industries to choose the appropriate RFID protocol for their needs.

cs.CR↗

arXiv2010

A Survey of RFID Authentication Protocols Based on Hash-Chain Method

Security and privacy are the inherent problems in RFID communications. There are several protocols have been proposed to overcome those problems. Hash chain is commonly employed by the protocols to improve security and privacy for RFID authentication. Although the protocols able to provide specific solution for RFID security and privacy problems, they fail to provide integrated solution. This article is a survey to closely observe those protocols in terms of its focus and limitations.

cs.CR↗