A support vector regression approach to detection in large-MIMO systems

We propose a support vector regression approach for symbol detection in large-MIMO systems employing spatial multiplexing. We explore the applicability of machine learning algorithms, in particular support vector machines, to address one of the recent research problem in communications.The machine learning capability is exploited to achieve fast detection in large dimension systems. The performance of the proposed method is compared with lattice reduction aided detection which is currently the popular choice and the improvement in terms of bit error rate is demonstrated. The sparse formulation of the problem matrix reduces the computational complexity and enables faster detection. The proposed detection algorithm is tailored to address both uncorrelated and correlated channel conditions as well.     

Multicast tree generation in networks with asymmetric links

We formulate the problem of multicast tree generation as one of computing a directed Steiner tree of minimal cost. In this context, we present a polynomial-time algorithm that provides for trade-off selection, using a single parameter κ, between the tree-cost (Steiner cost) and the run time efficiency. Further, the same algorithm may be used for delay optimization or tree-cost minimization simply by configuring the value of κ appropriately. We present theoretical and experimental analysis characterizing the problem and the performance of our algorithm. Theoretically, we show that it is highly unlikely that there exists a polynomial-time algorithm with a performance guarantee of constant times optimum cost, introduce metrics for measuring the asymmetry of graphs, and show that the worst-case cost of the tree produced by our algorithm is, at most, twice the optimum cost times the asymmetry, for two of these asymmetry metrics. For graphs with bounded asymmetry, this gives constant times optimum performance guarantee. We also show that three well-known algorithms for (undirected) Steiner trees are but particular cases of our algorithm. Our experimental study shows that operating at a low κ gives nearly best possible expected tree cost while maintaining acceptable run-time efficiency     

Broadcasting in multi-radio multi-channel wireless networks using simplicial complexes

We consider the broadcasting problem in multi-radio multi-channel ad hoc networks. The objective is to minimize the total cost of the network-wide broadcast, where the cost can be of any form that is summable over all the transmissions (e.g., the transmission and reception energy, the price for accessing a specific channel). Our technical approach is based on a simplicial complex model that allows us to capture the broadcast nature of the wireless medium and the heterogeneity across radios and channels. Specifically, we show that broadcasting in multi-radio multi-channel ad hoc networks can be formulated as a minimum spanning problem in simplicial complexes. We establish the NP-completeness of the minimum spanning problem and propose two approximation algorithms with order-optimal performance guarantee. The first approximation algorithm converts the minimum spanning problem in simplical complexes to a minimum connected set cover (MCSC) problem. The second algorithm converts it to a node-weighted Steiner tree problem under the classic graph model. These two algorithms offer tradeoffs between performance and time-complexity. In a broader context, this work appears to be the first that studies the minimum spanning problem in simplicial complexes and weighted MCSC problem.     

Development of an etchant for selectively etching TiWNxin the presence of electroplated 95%Pb-5%Sn solder

Shrinking die sizes and increasing I/O density is motivating the push toward flip chip packages. A flip chip interconnection system with a under bump metallurgy stack containing sputtered TiWN_X/sputtered Cu/electroplated Cu stud/electroplated 95%Pb-5%Sn was developed. An important step in the above process is the selective etching of the sputtered Cu bus layer and the TiWN_X barrier layer, in the presence of the Pb-Sn solder. The Cu bus layer was selectively etched using commercial etchants. However, no commercial etchants were available for selectively etching the TiWN_X layer, H₂O₂-NH₄OH based etching systems, popularly known as Standard Clean-1 cleaning solutions, have been extensively used to clean silicon wafers in front end wafer fabrication where only trace metal contamination exists. Since metals like lead, copper, titanium, tin and tungsten catalyze the heterogeneous decomposition of the peroxide, the unstable H₂O₂-NH₄OH based etching systems are rarely used to etch metal films. In this paper the development of a H ₂O₂-NH₄OH based etchant to selectively etch the sputtered TiWN_X films in the presence of electroplated 95%Pb-5%Sn solder bumps is discussed. A 2³ full factorial experiment with mid point was conducted to establish the etchant composition, as well as process temperature, that give satisfactory responses with respect to etch time, permissable undercut of the Cu stud (caused by the NH₄OH), and acceptable bump shape after reflow. Statistical analysis was used to understand the significant factors influencing the etch rate and undercut. An etchant containing 6% by volume of 30%-H₂O₂ and 0.75% by volume of 30%-NH₄OH operated at a temperature of 37°C was found to give satisfactory results     

An InGaP/GaAs Merged HBT-FET (BiFET) Technology and Applications to the Design of Handset Power Amplifiers

The last decade has seen GaAs HBTs emerge as the dominant technology in wireless handset power amplifiers. Modern application requirements and size limitations have driven industry leaders towards the co-integration of depletion mode n-FET and GaAs HBT. The merger of Bipolar and FET, or BiFET, gives an additional degree of freedom in the design of advanced power amplifiers independent of a silicon controller. This paper provides an overview of the various techniques that can be used to join the two device technologies and then shows how a merged epitaxial structure, where an FET is formed in the emitter layers of an HBT, combines functional versatility with the high volume manufacturability needed to supply millions of power amplifiers at low cost. A large-signal model of the FET structure is developed which takes into account the unique physics and geometries of the device, including voltage-dependant parameters and charges on all four electrical terminals. Specific handset applications that can benefit or be enabled by BiFET are presented, such as on-off switching, low voltage bias controllers , Auto-Bias power amplifiers, and bias circuits with low or no voltage reference. When npn-only bias circuitry is limited to low voltage reference levels, HBT power amplifiers with BiFET bias stages are shown to have superior RF performance to their npn-only counterparts.     

Electronic Granularity and the Work Function of Transparent Conducting ZnO:Al Thin Films

Controlling the efficiency of electron transport across oxide interfaces is essential for numerous emerging technologies including advanced photovoltaics and light emitting devices. This work illuminates the connections between granular structure, defect chemistry, and the work function of a technologically important transparent conductor, ZnO:Al. Visual evidence is provided for a model of grain boundary oxidation in the form of nanometer‐scale heterogeneity in the contact potential between grains and grain boundaries, a phenomenon referred to as electronic granularity. By correlating scanning probe data with photoemission spectroscopy we relate electronic granularity to defect chemistry and, importantly, account for the overall trends in work function. The resulting physical picture connects heterogeneity at the nanoscale to macroscopic properties, informs the design of transparent electrodes, and may be broadly relevant to granular oxide conductors.     

Face verification across age progression.

Human faces undergo considerable amounts of varialions with aging. While face recognition systems have been proven to be sensitive to factors such as illumination and pose, their sensitivity to facial aging effects is yet to be studied. How does age progression affect the similarity between a pair of face images of an individual? What is the confidence associated with establishing the identity between a pair of age separated face images? In this paper, we develop a Bayesian age difference classifier that classifies face images of individuals based on age differences and performs face verification across age progression. Further, we study the similarity of faces across age progression. Since age separated face images invariably differ in illumination and pose, we propose preprocessing methods for minimizing such variations. Experimental results using a database comprising of pairs of face images that were retrieved from the passports of 465 individuals are presented. The verification system for faces separated by as many as nine years, attains an equal error rate of 8.5%.     

A mathematical model for predicting the soil nitrogen status under varying fertilizer practices in a continuous cropping sequence

A mathematical model has been proposed for predicting the changes in soil nitrogen status due to continuous fertilization in a continuous cropping sequence. The model also enabled the prediction of the steady state of soil nitrogen for a specified fertilizer practice.The model was applied to six years nitrogen availability data of four fertilizer practices in finger millet-maize-cowpea sequence followed in the Long Term Fertilizer Experiments conducted at Tamil Nadu Agricultural University, Coimbatore, India. The agreement between the predicted soil nitrogen status by the model and the actuals was proved by employing reliability index.     

Pathlistings applied to data flow analysis

Summary In this paper the pathlisting mechanism is developed as a new tool useful in performing efficient data flow analysis of programs for a wide variety of problems. An algorithm using this tool for forward flow, code improvement problems is presented. It is shown that for all practical purposes this algorithm is linear in the size of the input which is, generally speaking, a reducible flow graph modeling the given program. Pathlistings generalize the nodelisting approach, introduced by Kennedy, for solving data flow problems. The efficiency of the pathlisting algorithm is due to the reuse of intermediate values and due to the fact that the cycles of a reducible flow graph can be ordered. Other advantages of the approach are also discussed.Work supported by National Science Foundation grant DCR73-00365-AO     

Effect of modified starch from sweet potato as a fat replacer on the quality of reduced fat ice creams

In the present study, application of different levels (0, 1 and 2%) of citric acid treated sweet potato starch as a fat replacer (FR) in a high (11% fat), medium (6% fat) and low-fat (1%) ice creams were investigated. Results indicated that hardness value tends to improve with the addition of 1% FR in all reduced fat ice creams. During 60 days of storage, a decreasing trend was noticed in overrun, acidity and hardness values of ice cream samples. Overall score of sensory conveyed that the reduced fat ice creams with 1% FR (medium fat ice cream and low fat ice cream) were found to be very acceptable by sensory panels with a similar physicochemical characteristics and acceptance as high-fat ice cream (control) at the end of storage period. Citric acid treated sweet potato starch proved to be a promising alternative as a fat replacer in the ice cream production.