Estimating the generalization error is one of the key ingredients of supervised learning since a good generalization error estimator can be used for model selection. An unbiased generalization error estimator called the subspace information criterion (SIC) is shown to be useful for model selection, but its range of application is limited to linear learning methods. In this paper, we extend SIC to be applicable to non-linear learning.
Takuji TACHIBANA Shoji KASAHARA
In this paper, we propose a new preemptive scheme with release message in optical burst switching (OBS) networks. In the proposed scheme, when a low priority burst is preempted at some intermediate node, two RELEASE messages are sent immediately from the intermediate node to both source and destination nodes (two-way release message transmission), and the RELEASE messages release the corresponding wavelengths for the preempted burst. We consider six wavelength selection rules for the preemption and evaluate the performances of the selection rules by simulations. Numerical examples show that our scheme utilizes wavelengths effectively and, with the optimal selection rule, can decrease the burst loss probability in a large-scale DWDM network.
Kazuya TSUKAMOTO Yoshiaki HORI Yuji OIE
A transport layer mobility management scheme for handling seamless handoffs between appropriate networks is presented. The future mobile environment will be characterized by multimodal connectivity with dynamic switching. Many technologies have been proposed to support host mobility across diverse wireless networks, and operate in various layers of the network architecture. Our major focus is on the transport protocol that recovers packets lost during handoffs and controls transmission speed to achieve efficient communication. Majority of the existing technologies can maintain the connection by updating the information of a single connection around a handoff. Moreover, none of the studies extensively examine the handoff latencies and focus how an appropriate network is selected, during the handoff. In this paper, we first extensively investigate the various handoff latencies and discuss the limited performance of existing technologies based on the single connection. We then propose a new scheme resolving the problems by the transport protocol enabling the adaptive selection of an appropriate interface based on communication condition among all available interfaces. Finally, we demonstrate that the proposed scheme promptly and reliably selects the appropriate interface, and achieves excellent goodput performance by comparing with the existing technologies.
The model selection for neural networks is an essential procedure to get not only high levels of generalization but also a compact data model. Especially in terms of getting the compact model, neural networks usually outperform other kinds of machine learning methods. Generally, models are selected by trial and error testing using whole learning samples given in advance. In many cases, however, it is difficult and time consuming to prepare whole learning samples in advance. To overcome these inconveniences, we propose a hybrid on-line learning system for a radial basis function (RBF) network that repeats quick learning of novel instances by rote during on-line periods (awake phases) and repeats pseudo rehearsal for model selection during out-of-service periods (sleep phases). We call this system Incremental Learning with Sleep (ILS). During sleep phases, the system basically stops the learning of novel instances, and during awake phases, the system responds quickly. We also extended the system so as to shorten the periodic sleep periods. Experimental results showed the system selects more compact data models than those selected by other machine learning systems.
This paper presents a novel threshold-based selection scheme to combine adaptive transmit antenna selection with an adaptive quadrature amplitude modulation (AQAM) for a spatial multiplexing (SM) multiple-input multiple-output (MIMO) system with linear receivers in practical uncorrelated and correlated channel conditions. The proposed scheme aims to maximize the average spectral efficiency (ASE) for a given bit error rate (BER) constraint and also to lower the hardware complexity. Our simulations are run on a general MIMO channel model, under the assumption that the channel state information (CSI) is known at the receiver and the adaptive control signaling can be perfectly fed back to the transmitter. We deploy the low rank-revealing QR (LRRQR) algorithm in transmit antenna subset selection. LRRQR is computationally less expensive than a singular value decomposition (SVD) based algorithm while the two algorithms achieve similar error rate performances. We show that both the conventional AQAM scheme (i.e., without adaptive transmit antenna selection) and the SM scheme perform poorly in a highly correlated channel environment. We demonstrate that our proposed scheme provides a well-behaved trade-off between the ASE and BER under various channel environments. The ASE (i.e., throughput) can be maximized with a proper choice of the channel quality threshold and AQAM mode switching threshold levels for a target BER.
Hierarchical Mobile IPv6 (HMIPv6) was designed to reduce the long handover time associated with MIPv6 by employing a hierarchy of Mobile Anchor Points (MAPs). However, the selection of MAP and its load status critically affect the system performance. Thus, in this paper, we introduce two novel dynamic MAP selection schemes for HMIPv6, that relieve overloaded MAPs as well as select a more suitable MAP according to each Mobile Node (MN)'s up-to-date mobility towards reducing inter-domain handover, resulting in saving the overall signaling cost. Further, we develop an analytical model of the average signaling cost for inter-domain handover to investigate the performance of another existing approach with its basis on IETF HMIPv6 as well as our schemes. We also perform simulations to evaluate the performance of our proposed schemes. It is shown via simulation and numerical results that the proposed dynamic MAP selection schemes can significantly reduce the number of inter-domain handovers and the average signaling cost as well as distribute load over multiple MAPs compared to the other existing approaches.
Tsuyoshi SADAKATA Yusuke MATSUNAGA
A Multi-Functional unit has several functions and these can be changed with a control signal. For High-Level Synthesis, using Multi-Functions units in operation chaining make it possible to obtaining the solution with the same number of control steps and less resources compared to that without them. This paper proposes an operation chaining method considering Multi-Functional units. The method formulates module selection, scheduling, and functional unit allocation with operation chaining as a 0/1 integer linear problem and obtains optimal solution with minimum number of control steps under area and clock-cycle type constraints. The first contribution of this paper is to propose the global search for operation chaining with Multi-Functional units having multiple outputs as well as with single output. The second contribution is to condier the area constraint as a resource constraint instead of the type and number of functional units. Experimental results show that chaining with Multi-Functional units is effective and the proposed method is useful to evaluate heuristic algorithms.
Haibo ZHENG Yongle WU Yunzhou LI Shidong ZHOU Jing WANG
In this letter, we propose a limited feedback precoding scheme based upon grassmannian beamforming and user selection for downlink multiuser MIMO systems. Conventional random beamforming scheme only enjoys significant performance gains with a large number of users, which limits its practical application. With proper codebook size the proposed scheme outperforms conventional random beamforming scheme when the number of users is small or moderate.
In this letter, we propose two different joint transmit and receive antenna subset selection schemes for multiple-input multiple-output (MIMO) systems on the basis of capacity maximization criterion. We assume that perfect channel state information (CSI) is known at the receiver but unknown to the transmitter. As the selection signaling is perfectly fed back to the transmitter, we propose a flexible two-step selection algorithm (TSSA) in practical MIMO channel scenarios. Computer simulations show that TSSA can maximize the capacity at low computation cost in most scenarios. It performs well in terms of capacity, computational complexity and flexibility. Furthermore, we propose a simplified algorithm based on the correlation matrix when the channel correlation information (CCI) is known to the transmitter. Simulation results show that the proposed correlation matrix based selection algorithm is only slightly inferior to an optimal selection algorithm.
Masatsune TAMURA Tatsuya MIZUTANI Takehiko KAGOSHIMA
We have previously developed a concatenative speech synthesizer based on the plural speech unit selection and fusion method that can synthesize stable and human-like speech. In this method, plural speech units for each speech segment are selected using a cost function and fused by averaging pitch-cycle waveforms. This method has a large computational cost, but some platforms require a speech synthesis system that can work within limited hardware resources. In this paper, we propose an offline unit fusion method that reduces the computational cost. In the proposed method, speech units are fused in advance to make a pre-fused speech unit database. At synthesis time, a speech unit for each segment is selected from the pre-fused speech unit database and the speech waveform is synthesized by applying prosodic modification and concatenation without the computationally expensive unit fusion process. We compared several algorithms for constructing the pre-fused speech unit database. From the subjective and objective evaluations, the effectiveness of the proposed method is confirmed by the results that the quality of synthetic speech of the offline unit fusion method with 100 MB database is close to that of the online unit fusion method with 93 MB JP database and is slightly lower to that of the 390 MB US database, while the computational time is reduced by 80%. We also show that the frequency-weighted VQ-based method is effective for construction of the pre-fused speech unit database.
Masato MIZUKAMI Yoshitada KATAGIRI
We propose and demonstrate wavelength-selectable filters available for 32 WDM channels using a micro-mechanically movable mechanism with miniaturized voice-coil motors (VCMs). A simple straight geometry with a staggered configuration is used to densely pack 32 in/out moving elements into a small space of 452411 mm. The elements are precisely arranged along a collimated beam between fiber facets to provide flat-top passbands centered at ITU-T grids while maintaining small total insertion losses of less than 2.5 dB for all elements. The driving condition of the VCMs is also optimized for quick dynamic response with typical settling time of less than 10 ms. A repetition test 106 repetitions per element showed good wavelength reproducibility to an accuracy of below 0.1 nm, indicating the switches are feasible for practical system equipped with reconfigurable functionality for the next generation of optical networks.
Shunsuke SAITO Yasuyuki TANAKA Mitsunobu KUNISHI Yoshifumi NISHIDA Fumio TERAOKA
Recently, the number of multi-homed hosts is getting large, which are equipped with multiple network interfaces to support multiple IP addresses. Although there are several proposals that aim at bandwidth aggregation for multi-homed hosts, few of them support mobility. This paper proposes a new framework called AMS: Aggregate-bandwidth Multi-homing Support. AMS provides functions of not only bandwidth aggregation but also mobility by responding to the changes of the number of connections during communication without the support of underlying infrastructure. To achieve efficient data transmission, AMS introduces a function called address pairs selection to select an optimal combination of addresses of the peer nodes. We implemented AMS in the kernel of NetBSD and evaluated it in our test network, in which dummynet was used to control bandwidth and delay. The measured results showed that AMS achieved ideal bandwidth aggregation in three TCP connections by selecting optimal address pairs.
We investigate selection transmit multi-input multi-output systems where only a single transmit antenna is selected for the transmission and multiple receive antennas are employed for maximal ratio combining. Antenna selection is performed by a generalized selection criterion based on the ordinal number of the strength of the received signal-to-noise ratio.
Anand PAUL Jhing-Fa WANG Jia-Ching WANG An-Chao TSAI Jang-Ting CHEN
This paper introduces a block based motion estimation algorithm based on projection with adaptive window size selection. The blocks cannot match well if their corresponding 1D projection does not match well, with this as foundation 2D block matching problem is translated to a simpler 1D matching, which eliminates majority of potential pixel participation. This projection method is combined with adaptive window size selection in which, appropriate search window for each block is determined on the basis of motion vectors and prediction errors obtained for the previous block, which makes this novel method several times faster than exhaustive search with negligible performance degradation. Encoding QCIF size video by the proposed method results in reduction of computational complexity of motion estimation by roughly 45% and over all encoding by 23%, while maintaining image/video quality.
Masanori HARIYAMA Shigeo YAMADERA Michitaka KAMEYAMA
This paper presents a design method to minimize energy of both functional units (FUs) and an interconnection network between FUs. To reduce complexity of the interconnection network, data transfers between FUs are classified according to FU types of operations in a data flow graph. The basic idea behind reducing the complexity of the interconnection network is that the interconnection resource can be shared among data transfers with the same FU type of a source node and the same FU type of a destination node. Moreover, an efficient method based on a genetic algorithm is presented.
Nandar LYNN Osamu TAKYU Riaz ESMAILZADEH Masao NAKAGAWA
In this paper, we evaluate the performance of asymmetric Time Division Duplex (TDD) system that employs Adaptive Modulation and Coding (AMC) and Hybrid ARQ, with consideration of the effect of control delays in TDD. Channel reciprocity characteristic in TDD allows utilization of open loop channel estimation to choose appropriate modulation and coding scheme (MCS) level for AMC. However, control delay in AMC and HARQ depends on TDD time slot allocation formats. Large control delay in AMC will result in false MCS selection due to the poor channel correlation between measured channel state from the received signals and instantaneous channel state of actual transmission with the MCS selected based on the measured channel state. We present an analytical approach to calculate the probability of MCS level selection error in different channel conditions for different asymmetric time slot allocations. From the theoretical and simulation results, it is shown that the instantaneous throughput per slot depends not only on maximum Doppler frequency but also on asymmetric slot allocations. Average delay time that yields error free packet reception in the downlink increases as the number of continuous downlink slots increases.
Tadashi OKUBO Ryo MOCHIZUKI Tetsunori KOBAYASHI
We propose a hybrid voice conversion method which employs a combination of techniques using HMM-based unit selection and spectrum generation. In the proposed method, the HMM-based unit selection selects the most likely unit for the required phoneme context from the target speaker's corpus when candidates of the target unit exist in the corpus. Unit selection is performed based on the sequence of the spectral probability distribution obtained from the adapted HMMs. On the other hand, when a target unit does not exist in a corpus, a target waveform is generated from the adapted HMM sequence by maximizing the spectral likelihood. The proposed method also employs the HMM in which the spectral probability distribution is adjusted to the target prosody using the weight defined by the prosodic probability of each distribution. To show the effectiveness of the proposed method, sound quality and speaker individuality tests were conducted. The results revealed that the proposed method could produce high-quality speech and individuality of the synthesized sound was more similar to the target speaker compared to conventional methods.
Shuangfeng HAN Shidong ZHOU Ming ZHAO Jing WANG Kyung PARK
Aiming to optimally transmit space-time block codes (STBCs) over distributed antennas (DAs), this paper examines downlink transmit antenna subset selection with power allocation for STBCs in non-ergodic Rayleigh fading channels with receive antenna correlations. Closed-form outage probability is first derived, which is a function of data rate, rate of STBCs, transmit power, large-scale fading (shadowing and path loss), power allocation weights to each DA and receive antenna correlation. However, achieving the optimal power allocation solution is computationally demanding and the use of sub-optimal techniques is necessitated. Assuming feedback of eigenvalues of transmit and receive antenna correlation matrix at the transmitter and accurate channel state information (CSI) at the receiver, an antenna subset selection with sub-optimal power allocation scheme is proposed, whose performance approaches optimal. The effectiveness of this sub-optimal method has been demonstrated by numerical results.
Haibo ZHENG Xiang CHEN Shidong ZHOU Jing WANG Yongxing ZHOU James Sungjin KIM
In this letter, we propose an efficient user selection algorithm aiming to select users with less spatially correlation and meet the user number limit of zero-forcing beamforming in downlink multiuser MIMO systems. This algorithm yields a considerable complexity reduction with only a small loss in performance and it only needs partial users' CSI feedback. Coupled with the algorithm, a null space updating method in O(K2) time and a modified proportional fair scheduling algorithm are also proposed.
Masashi SUGIYAMA Keisuke SAKURAI
For obtaining a higher level of generalization capability in supervised learning, model parameters should be optimized, i.e., they should be determined in such a way that the generalization error is minimized. However, since the generalization error is inaccessible in practice, model parameters are usually determined in such a way that an estimate of the generalization error is minimized. A standard procedure for model parameter optimization is to first prepare a finite set of candidates of model parameter values, estimate the generalization error for each candidate, and then choose the best one from the candidates. If the number of candidates is increased in this procedure, the optimization quality may be improved. However, this in turn increases the computational cost. In this paper, we give methods for analytically finding the optimal model parameter value from a set of infinitely many candidates. This maximally enhances the optimization quality while the computational cost is kept reasonable.