Estimating variable structure and dependence in multitask learning via gradients

We consider the problem of hierarchical or multitask modeling where we simultaneously learn the regression function and the underlying geometry and dependence between variables. We demonstrate how the gradients of the multiple related regression functions over the tasks allow for dimension reduction and inference of dependencies across tasks jointly and for each task individually. We provide Tikhonov regularization algorithms for both classification and regression that are efficient and robust for high-dimensional data, and a mechanism for incorporating a priori knowledge of task (dis)similarity into this framework. The utility of this method is illustrated on simulated and real data.     

Processing of duplex stainless steel by WEDM

This study investigates the manufacturing process of 2205 duplex stainless steel by wire electrical discharge machining where the effects of pulse-on time (PONT), wire tension and pulse-off time (POFT) on surface finish, kerf width, and material removal rate (MRR). It was found that the kerf width was unchanged with the change of PONT at long pulse-of time and higher wire tension. However, it decreased initially and then increased due to the rise of PONT at low values of wire tension and POFT. Low wire tension and PONT, POFT and contributed towards widest kerf. Longer PONT increased MRR due to higher machining/processing speed. Lower wire tension and shorter POFT increased MRR more than that of higher wire tension and POFT. Craters and recast layer were on the machined surfaces at all machining conditions. Increased PONT raised surface roughness at the lower POFT and tension in the wire. The surface finish at high wire tension and longer PONT is always better than that at smaller PONT and lower tension in the wire. The microstructure underneath the recast layer remains unchanged and the failure of wire electrode occurred at higher wire tension, longer PONT and shorter POFT.     

Exploring the limits of stress-enhanced hole mobility

Hole mobility is found to more than double in fabricated p-MOSFETs with SiGe source/drain due to longitudinal compressive stress in the channel exceeding 1 GPa. The maximum observed low-field mobility enhancement is 140% at a simulated stress level of 1.45 GPa. The mobility enhancement is approximately linear with stress at moderate levels but becomes super-linear above 1 GPa. An important consequence of this behavior is that for moderate stress levels, an average channel stress can be used to estimate the performance of transistors with a nonuniform stress distribution across the channel width. Two alternative approaches to model stress-enhanced hole mobility are suggested. Analysis of the physical effects behind the experimental observations reveals the relative roles of band repopulation and mass modulation. In addition, previously published wafer bending experiments with compressive stress levels below 400 MPa are used to implicitly verify the accuracy of the stress simulations.     

New chemical methods for the deposition of Cu1·8S and TlSe thin film

New chemical methods for the deposition of thin film of Cu_1·8S and TlSe have been developed. The deposition of Cu_1·8S thin film has been performed by thiourea, ammonia and Cu²⁺ ions at room temperature, while TlSe thin films are obtained from triethanolamine as complexing agent, ammonia, sodium selenosulphate solution and Tl¹⁺ ions at room temperature. The electrical resistance, mobility, carrier concentration and optical band gap have been measured.     

Generative model based robotic grasp pose prediction with limited dataset

Applied Intelligence - In the present investigation, we propose an architecture which we name as Generative Inception Neural Network (GI-NNet), capable of predicting antipodal robotic grasps...     

Retraction Note to: Sintered Aluminum–Graphene Nano-Bio Composite Materials for the Medical Application

Powder Metallurgy and Metal Ceramics -     

Optimizing dimensional accuracy of titanium alloy features produced by wire electrical discharge machining

This study investigates geometrical errors such as cylindricity, circularity and diametral errors of a feature (a hole) produced from wire electrical discharge machining of Ti6Al4V alloy where tension in wire, pulse on time, and flushing pressure are varied. Pareto analysis of variance (ANOVA), Taguchi design of experiment (DoE), and traditional analysis estimate the influence of variables on errors of holes. It was noted that flushing pressure is the most significant factor with individual contributions of 31.02%, 49.5% and 37.84% to circularity, cylindricity, and diametral errors, respectively. The circularity error of holes decreases as the flushing pressure and tension in wire rise, but decreases with the rise of pulse on time. The cylindricity error decreased with the increase of wire tension, flushing pressure and pulse on time. The absolute diametral error reduced as the pulse on time and tension in wire raised, but it raised with the rise of pulse on time. All these trends are associated with the influence of tension in wire on the flexibility of wire, the dependence of heat generation and dissipation on pulse on time, and ability of the flushing pressure to control the cooling, as well as debris removal from the machining zone.     

Burr formation during drilling of mild steel at different machining conditions

This paper investigates burr generation while drilling of mild steel grade 350. The influences of feed, speed and point angle on burr height, thrust force, torque and chip ratio are investigated to correlate with burr height. It was found that, the burr height reduces gradually with the rise of speed at minimum feed and point angle. At maximum feed and point angle, initially the height rises with speed and then reduces as speed rises further. The maximum burr is 720 μm at 584 rpm. At maximum point angle and speed, the height rises initially and then reduces as the feed rises where the trend is opposite at minimum feed and speed. The maximum burr is 1223.15 μm at 0.25 mm/rev feed. The variation of burr height with point angle is similar to that with the variation of feed. The maximum burr is 1230 μm at 125°point angle. The trends of thrust forces, torques and chip ration with the variation of different parameters are not similar to that of burr height in most of the cases. The complex interaction between strain hardening and thermal softening plays the main role in burr formation for the considered material.