Summary The automobile transmission is the core component of the automobile transmission system. Extracting effective fault feature information from a complex mechanical device like a car transmission is the key to fault diagnosis. In this paper, the vibration signals of the normal transmission and different faulty transmissions are collected through experiments, and the weak fault feature extraction of automobile transmission gears is based on the combination of Teager energy operator demodulation method and empirical mode decomposition method. Finally, the theory of SVM and its application in fault diagnosis are introduced. Based on the cross‐validation method, the grid search method and particle swarm optimization algorithm are used to search and optimize the important parameters in the SVM model to improve the performance of the SVM classifier.

A large number of new buyers are often acquired by merchants during promotions. However, many of the attracted buyers are one-time deal hunters, and the promotions may have little long-lasting impact on sales. It is important for merchants to identify who can be converted to regular loyal buyers and then target them to reduce promotion cost and increase the return on investment (ROI). At International Joint Conferences on Artificial Intelligence (IJCAI) 2015, Alibaba hosted an international competition for repeat buyer prediction based on the sales data of the ``Double 11" shopping event in 2014 at Tmall.com. We won the first place at stage 1 of the competition out of 753 teams. In this paper, we present our winning solution, which consists of comprehensive feature engineering and model training. We created profiles for users, merchants, brands, categories, items and their interactions via extensive feature engineering. These profiles are not only useful for this particular prediction task, but can also be used for other important tasks in e-commerce, such as customer segmentation, product recommendation, and customer base augmentation for brands. Feature engineering is often the most important factor for the success of a prediction task, but not much work can be found in the literature on feature engineering for prediction tasks in e-commerce. Our work provides some useful hints and insights for data science practitioners in e-commerce.

In this paper, we propose an adaptive p-persistent based IEEE 802.11 medium access control (MAC) scheme in a heterogeneous WLAN. Quality of Service in a DCF based heterogeneous WLAN is a challenging task due to the lack of centralized scheduling capability. The proposed scheme can maximize the total channel throughput and provide the service differentiation among different traffic stations. This is achieved by updating the transmission probability for each station in a timely manner based on the real-time network measurements. The simulation results show that the scheme can quickly adapt the station transmission probabilities to the desirable values in order to achieve maximum throughput and QoS provisioning in a dynamic WLAN environment.

With the explosively increasing demand of multimedia applications in wireless local area networks (WLAN), Quality of Service (QoS) provisioning has become an important issue. IEEE 802.11 WLAN is the most popular WLAN technology today. In this paper, we propose an adaptive p-persistent based IEEE 802.11 medium access control (MAC) scheme in WLAN. The proposed scheme can maximize the total channel throughput and also can provide the service differentiation to multiple traffic classes. This is achieved by updating the transmission probability for each station in a timely manner based on the network measurements. The simulation results show the scheme can quickly adapt the station transmission probabilities to the desirable values to achieve maximum throughput and QoS provisioning in a dynamic WLAN environment. The simulation results also match well with the theoretical analysis.

The ergodic capacity of Ricean MIMO (multiple-input multiple-output) fading channels is investigated in this paper, based on a virtual (Fourier) channel representation. Coherent reception is assumed at the receiver. The Ricean statistics are due to a line-of-sight (LOS) path joining the transmitter and receiver. In the virtual domain, this LOS path corresponds to a single entry in the virtual channel matrix. The optimal input distribution of the correlated Ricean MIMO channel in the virtual domain has a diagonal covariance matrix structure, the same as in the Rayleigh fading MIMO case. Our method is used to get a tight lower bound and upper bound on the ergodic capacity. Further, the same analysis is used to obtain an approximate optimal input distribution analytically, in place of the usual Monte Carlo based numerical optimization methods.

siRNAs for gene silencing studies have a broad range of applications. Studies have addressed the design of siRNAs based on the target sequence patterns, such as the sequence AA around the ends of the siRNA or the avoidance of GGG in siRNA sequences. However, information from just these two patterns is not sufficient for effectively designing siRNA. Nine sequence patterns were identified as highly significantly correlated motifs in studies on oligonucleotides antisense activities. The correlation between the sequence patterns of the siRNA and the gene silencing ratio (GSR) by siRNA has not been described. Understanding this relationship may guide one to a systematic design of the most effective siRNAs. We designed a computer program to calculate the sequence patterns from our siRNA silencing database. Four statistical methods were used to identify the frequency and the distribution of the highly significant sequence patterns from the 1344 combinations for the three, four, and five nucleotide sequences. For the first time, 52 sequence patterns were identified as the highly significant patterns by ANOVA and Pearson correlation analyses with the GSR. These 52 patterns were further clustered into 18 groups. Information from this study may help in the optimized design of siRNA.

siRNA is a powerful technique for gene silencing. However, it is still not clear how to accurately determine an effective gene targeting region for siRNA. It is also controversial if the targeted mRNA secondary structure would have a significant impact on the gene silencing by the siRNA. Understanding the relationship between the secondary structure and gene silencing by siRNA could help in high throughput function-based genetic screens and for development of therapeutic tools. Our hypothesis is that the selection of a single-stranded region as the target region should increase the efficiency of siRNA silencing. For the first time, a formula and a computer program were developed to select the single-stranded regions for siRNA silencing. We statistically analyzed quantitative relationships between gene silencing by the siRNA and ten parameters from published siRNA sequences. The four nonlinear significant parameters for designing siRNA are the RNA single-stranded index (SSi), a new parameter that identifies the target position and its related secondary structures, the GC content, /spl Delta/G free energy of the siRNA, and the 5'GC content of the first four bases of the antisense siRNA. The nonlinear complexity suggested that one should not use simple criteria such as a GC content of around 30 to 50% or the absence of the sequence pattern GGG for designing siRNA. Our results indicate that the targeted mRNA secondary structure is a statistically significant determinant in gene silencing and should be used to optimize RNAi technology.

Existing objective speech quality measurement algorithms still fall short of the measurement accuracy that can be obtained from subjective listening tests. We propose an approach that uses statistical data mining techniques to improve the accuracy of auditory-model based quality measurement algorithms. We present the design of a novel measurement algorithm using the multivariate adaptive regression splines (MARS) method. A large set of speech distortion features is first created. MARS is used to find a small set of features that provide the best estimate ("model") of speech quality. One appeal of the approach is that the model size can scale with the amount of speech data available for learning. In our simulations, the new algorithm furnishes significant performance improvement over PESQ (perceptual evaluation of speech quality).

We propose a stable and reduced-complexity detection method for the Bell Labs layered space-time (BLAST) coding system. The existing iterative nulling and cancellation algorithm for BLAST has high computational complexity and requires repeated matrix pseudo-inverse calculation which may lead to numerical instability. A square-root algorithm was proposed by other researchers to reduce complexity and improve numerical stability. In this paper, to further reduce complexity, we modify the decorrelating decision-feedback CDMA multiuser detection method and apply it to BLAST. Similar to the square-root algorithm, numerical stable unitary transformations are performed on the Cholesky-decomposed matrices to reorder the detection and cancellation steps. However, our method exploits a symmetry property in the triangularization process, which may further improve the numerical stability and reduce computational complexity over that of the square-root algorithm. In the simulation results, the impact of the reordering on performance is demonstrated.

We investigate a new robust multiuser signal detector for asynchronous code-division multiple-access uplink channels under delay mismatch. We first formulate a robust decorrelating detector by dividing each user into two virtual users with rectangular chip pulse shapes. To increase the system capacity, a multistage version of the robust decorrelating detector is derived, which can achieve capacity of up to M/(M+1) of the spreading factor, where M is the observation block length. We further propose a robust successive interference cancellation (SIC) implementation. The proposed robust SIC detector adds only a residual error estimation procedure onto the standard SIC detector, so its computational complexity is of the same order of that of the SIC. Performance is investigated via analysis and simulation. Computer simulation results showed that our proposed robust SIC detector outperforms the conventional decorrelating detector when delay estimation error is present, and its performance is close to that of the decorrelating detector with perfect time-delay information. Finally, we generalize the robust SIC detector to the case of nonrectangular chip pulse shapes.

Successive interference cancellation (SIC) is a low-complexity multiuser detection method for DS/CDMA systems. If the channel is AWGN or slowly varying, averaging the channel amplitude estimates over multiple bits can greatly improve performance. However, the error propagation associated with a hard-limiter decision rule prevents SIC from achieving the performance improvement from averaging. We propose a new soft-decision rule to be used with amplitude averaging, which combines linear and hard decision rules. A performance within 0.4 dB of the single-user bound is obtained. We also made the above soft-decision SIC robust to time delay errors.

We investigate an improved multiuser detector for asynchronous CDMA uplink channels when there exist timing delay estimation errors. We first formulate a robust decorrelating detector with a capacity limit of 50%. To increase the system capacity, we further propose a robust successive interference cancellation (SIC) implementation. The proposed robust SIC detector adds only a residual error estimation procedure onto the standard SIC detector. Computer simulation results show that its performance and capacity are very close to those of the ideal decorrelating detector which assumes perfect knowledge of time delay.