Reinforcement Learning Based Quantization Strategy Optimal Assignment Algorithm for Mixed Precision

The quantization algorithm compresses the original network by reducing the numerical bit width of the model,which improves the computation speed. Because different layers have different redundancy and sensitivity to databit width. Reducing the data bit width will result in a loss of accuracy. Therefore, it is difficult to determinethe optimal bit width for different parts of the network with guaranteed accuracy. Mixed precision quantizationcan effectively reduce the amount of computation while keeping the model accuracy basically unchanged. In thispaper, a hardware-aware mixed precision quantization strategy optimal assignment algorithm adapted to low bitwidth is proposed, and reinforcement learning is used to automatically predict the mixed precision that meets theconstraints of hardware resources. In the state-space design, the standard deviation of weights is used to measurethe distribution difference of data, the execution speed feedback of simulated neural network accelerator inferenceis used as the environment to limit the action space of the agent, and the accuracy of the quantization model afterretraining is used as the reward function to guide the agent to carry out deep reinforcement learning training. Theexperimental results show that the proposed method obtains a suitable model layer-by-layer quantization strategyunder the condition that the computational resources are satisfied, and themodel accuracy is effectively improved.The proposed method has strong intelligence and certain universality and has strong application potential in thefield of mixed precision quantization and embedded neural network model deployment.

Performance evaluation of the fixed relay in OFDMA systems

Performance of fixed relays in orthogonal frequency division multiple access （0FDMA） systems, especially the coverage and capacity performance at the cell edge, is evaluated in this paper. Two methods, theoretical analysis and calculation and Monte Carlo simulation, are used for the evaluations. By theoreti- cal analysis and calculation, frequency efficiency equation of a relay is introduced and numerical results are calculated. Monte Carlo simulation results also verify that the calculation method is reasonable. The evaluation shows that a relay can increase system performance to a certain level if it is designed appropriately, otherwise it will be harmful for the system performance, even to reduce it.

Study of IMT-advanced heterogeneous network

Referring to research on the Heterogeneous Network （Het-Net） application scenario and technique characters in IMT-Advaneed （The 4th Generation Mobile Communications） cellular system, this paper provides further analysis on main technique aspects of Heterogeneous Network, discussion on interference issue due to multi-layer building by access points and their corresponding solutions from standardization and engineering implementation. The proposed solution can effectively solve the interference problem in IMT-advanced Het-Net, and also improves the networking performance dramaticaUy for future mobile communication systems.

ZTE Partners with KPN Group Belgium to Deploy Packet-Switched Core Network

27 August 2012--ZTE Corporation has signed a deal on a packet-switched core network （CN） for KPN Group Belgium （KPNGB）. KPNGB will deploy ZTE＇s packet-switched CN equipment, which supports unified radio access. The contract is the second of its kind between ZTE and KPNfollows from a construction project with KPN Germany （E-Plus） that was completed in September 2010.

ZTE Plans to Deploy Colombia's First Mobile WiMAX Network by the End of 2008

ZTE Corporation announced on August19,2008 that it had clinched an exclusive bid from Colombia’s leading operator

Techniques for Determining the Geographic Location of IP Addresses in ISP Topology Measurement

A brief survey on the state-of-the-art research of determining geographic location of IP addresses is presented. The problem of determining the geographic location of routers in Internet Service Provider （ISP） topology measurement is discussed when there is inadequate information such as domain names that could be used. Nine empirical inference rules are provided, and they are respectively （1） rule of mutual inference, （2） rule of locality, （3） rule of ping-pong assignment, （4） rule of bounding from both sides, （5） rule of preferential exit deny, （6） rule of uureachable/timeout, （7） rule of relay hop assignment, （8） rule of following majority, and （9） rule of validity checking based on interface-finding. In totally 2,563 discovered router interfaces of a national ISP topology, only 6.4% of them can be located by their corresponding domain names. In contrast, after exercising these nine empirical inference rules, 38% of them have been located. Two methods have mainly been employed to evaluate the effectiveness of these inference rules. One is to compare the measured topology graph with the graph published by the corresponding ISP. The other is to contact the administrator of the corresponding ISP for the verification of IP address locations of some key routers. The conformity between the locations inferred by the rules and those determined by domain names as well as those determined by whois information is also examined. Experimental results show that these empirical inference rules play an important role in determining the geographic location of routers in ISP topology measurement.

Special issue on future network:software-defined networking

Computer networks have to support an everincreasing array of applications,ranging from cloud computing in datacenters to Internet access for users.In order to meet the various demands,a large number of network devices running different protocols are designed and deployed in networks.