College of Engineering, Pune

About: College of Engineering, Pune is a based out in . It is known for research contribution in the topics: Computer science & Sliding mode control. The organization has 4264 authors who have published 3492 publications receiving 19371 citations. The organization is also known as: COEP.

Vehicle-Type Classification Using Customized Fuzzy Convolutional Neural Network

Abstract: Vehicle-type classification is fast becoming a popular domain of interest due to its various application areas ranging from traffic surveillance to autonomous navigation. In this paper, customized fuzzy convolutional neural network (CFCNN) is proposed for the classification of vehicle types by incorporating a customized convolutional neural network (CCNN) with a fuzzy hypersphere neural network (FHSNN). The two pivotal factors for the improvement of the CFCNN learning algorithm are: first, the ability to extract dominant hidden features from vehicle front raw images using CCNN and second, fuzzy hypersphere membership function used for pattern classification. The proposed model's performance evaluated using the BIT-vehicle standard dataset, which consists of high-resolution 9850 images of various five types of vehicles, and the testing accuracy obtained 94.26%, which is exceptionally better in comparison to the existing algorithms.

Numerical Analysis of New Airfoils for Small Wind Turbine Blade

Abstract: Wind energy is the major renewable energy source available easily in abundance. It has become more popular than other renewable energy sources because of its compatible cost with conventional energy sources. The demand of small wind turbine in the market with a reliable, durable and more efficient technology is increasing. Blade is the key element of wind turbine. Performance of wind turbine blade mainly depends on its airfoil. In the present work, new airfoil is developed with the aim of performance improvement of wind turbine. Two different airfoils are designed for root sections and tip sections. Blade Element Momentum theory is used in designing of the small wind turbine blade. The numerical analysis of blade airfoil is carried out by Q-Blade software. The results obtained are compared with NACA 2412 and SG 6042 airfoils. In addition, performances of blades made with individual airfoil are compared by using Q-Blade software. These comparisons are made by keeping the length of the blade and chord lengths of various airfoils constant. The new airfoils have shown better performance than NACA 2412 and SG 6042 airfoils. Keywords: Q-Blade, wind turbine blades, airfoil Cite this Article: Kale SA, Birajdar MR, Sapali SN, Numerical Analysis of New Airfoils for Small Wind Turbine Blade. Journal of Alternate Energy Sources & Technologies. 2015; 6(1): 1–6p.

Assessment of Autoencoder Architectures for Data Representation

Abstract: Efficient representation learning of data distribution is part and parcel of successful execution of any machine learning based model. Autoencoders are good at learning the representation of data with lower dimensions. Traditionally, autoencoders have been widely used for data compression in order to represent the structural data. Data compression is one of the most important tasks in applications based on Computer Vision, Information Retrieval, Natural Language Processing, etc. The aim of data compression is to convert the input data into smaller representation retaining the quality of input data. Many lossy and lossless data compression techniques like Flate/deflate compression, Lempel–Ziv–Welch compression, Huffman compression, Run-length encoding compression, JPEG compression are available. Similarly, autoencoders are unsupervised neural networks used for representing the structural data by data compression. Due to wide availability of high-end processing chips and large datasets, deep learning has gained a lot attention from academia, industries and research centers to solve multitude of problems. Considering the state-of-the-art literature, autoencoders are widely used architectures in many deep learning applications for representation and manifold learning and serve as popular option for dimensionality reduction. Therefore, this chapter aims to shed light upon applicability of variants of autoencoders to multiple application domains. In this chapter, basic architecture and variants of autoencoder viz. Convolutional autoencoder, Variational autoencoder, Sparse autoencoder, stacked autoencoder, Deep autoencoder, to name a few, have been thoroughly studied. How the layer size and depth of deep autoencoder model affect the overall performance of the system has also been discussed. We also outlined the suitability of various autoencoder architectures to different application areas. This would help the research community to choose the suitable autoencoder architecture for the problem to be solved.

Employing stacked ensemble approach for time series forecasting

Abstract: This manuscript presents a novel stack-based multi-level ensemble model to forecast the future incidences of conjunctivitis disease. Besides predicting the frequency of conjunctivitis, the proposed model also enhances accuracy through the use of the ensemble model. A stacked multi-level ensemble model based on Auto-ARIMA (Autoregressive Integrated Moving Average), NNAR (Neural Network Autoregression), ETS (Exponential Smoothing), HW (Holt Winter) is proposed and applied on the dataset. Predictive analysis is carried out on the collected dataset and further evaluated for various performance measures. The result shows that the various error metrics of the proposed ensemble is decreased significantly. Considering the RMSE (Root Mean Square Error) error values, for instance, are reduced by 39.23%, 9.11%, 19.48%, and 17.14% in comparison to Auto-ARIMA, NNAR, ETS, and HW model in that order. This research concludes that the accuracy of the forecasting of diseases can be significantly increased by applying the proposed stack generalization ensemble model as it minimizes the prediction error and hence provides better prediction trends as compared to Auto-ARIMA, NNAR, ETS, and HW model applied discretely.