Many approaches in generalized zero-shot learning rely on cross-modal mapping between the image feature space and the class embedding space. As labeled images are expensive, one direction is to augment the dataset by generating either images or image features. However, the former misses fine-grained details and the latter requires learning a mapping associated with class embeddings. In this work, we take feature generation one step further and propose a model where a shared latent space of image features and class embeddings is learned by modality-specific aligned variational autoencoders. This leaves us with the required discriminative information about the image and classes in the latent features, on which we train a softmax classifier.The key to our approach is that we align the distributions learned from images and from side-information to construct latent features that contain the essential multi-modal information associated with unseen classes. We evaluate our learned latent features on several benchmark datasets, i.e. CUB, SUN, AWA1 and AWA2, and establish a new state of the art on generalized zero-shot as well as on few-shot learning. Moreover, our results on ImageNet with various zero-shot splits show that our latent features generalize well in large-scale settings.

No abstract

No abstract

No abstract

No abstract

No abstract

SUMMARY An optimal controller has been designed for automatic generation control (AGC) of two‐area restructured power system. The performance of the controller has been studied for two cases: (a) when the tie line connecting the two areas consists of parallel EHVAC and HVDC lines and (b) when HVDC link is absent and only ac tie line acts as interconnection. The comparison has been done for the two cases in terms of time taken to damp out oscillations of frequency in each area, tie‐line power exchange between the two areas and power generated by each generator when there is change in load. The simulation results indicate that control performance improves when HVDC link is connected in parallel with EHVAC tie line and output of the dc block is considered as additional state for the optimal controller. Detailed simulation studies were carried out for load demand by DISCOs as per cotracted condition as well as for bilateral contract violation condition. Optimal controller gives better dynamic response compared to conventional PI controller. AGC dynamic response is further improved by using combined optimal fuzzy controller. Particle swarm optimization (PSO) technique has been used for tuning the fuzzy controller. The results of the two area thermal–thermal system indicate better control performance in terms of peak overshoot and settling time by PSO tuned combined optimal fuzzy controller as compared to the un‐tuned controller. Copyright © 2011 John Wiley & Sons, Ltd.

Power flow study is the initial step which provides voltage magnitudes, phase angles, active and reactive power flows at respective buses under normal operating conditions. It helps in analyzing the current state of the power system and effective alternative options for expanding the existing system in order to meet with the increasing load demand. Contingency Analysis is also a major part of study for reliable and planned operation of power system. It is very significant function in modern Energy Management Systems. The objective of contingency analysis is to give operator the information about the static security. There are several factors which may lead to the contingency in power system, for example line outage, transformer outage, generator outage and overloads resulting the extreme situations such as voltage collapse, over loads in other branches and/or sudden system voltage rise or drop. Contingency analysis is used to calculate parameters violations. In this paper, maximum loading parameter is calculated and contingency status of Western System Coordinating Council 3 Machine, 9 Bus test system is done using PSAT toolbox in MATLAB.