We present an efficient foveal framework to perform object detection. A scale normalized image pyramid (SNIP) is generated that, like human vision, only attends to objects within a fixed size range at different scales. Such a restriction of objects’ size during training affords better learning of object-sensitive filters, and therefore, results in better accuracy. However, the use of an image pyramid increases the computational cost. Hence, we propose an efficient spatial sub-sampling scheme which only operates on fixed-size sub-regions likely to contain objects (as object locations are known during training). The resulting approach, referred to as Scale Normalized Image Pyramid with Efficient Resampling or SNIPER, yields up to 3× speed-up during training. Unfortunately, as object locations are unknown during inference, the entire image pyramid still needs processing. To this end, we adopt a coarse-to-fine approach, and predict the locations and extent of object-like regions which will be processed in successive scales of the image pyramid. Intuitively, it's akin to our active human-vision that first skims over the field-of-view to spot interesting regions for further processing and only recognizes objects at the right resolution. The resulting algorithm is referred to as AutoFocus and results in a 2.5-5× speed-up during inference when used with SNIP. Code: https://github.com/mahyarnajibi/SNIPER .
For classification of High Dimensional data, feature selection is the most important step for obtaining optimal result with respect to processing power required and time taken. Feature selection is a method by which the most relevant feature is selected from a set of features containing redundant and irrelevant features thereby reducing the load on the classification algorithm. This paper proposes an implementation of this method in a two tier structure. In the first step, a high ranking feature is selected using the well-known filter based algorithm - Fisher Score. This algorithm selects the relevant feature from the feature set based on a preset threshold. The second step generates a cluster of redundant features utilizing MST (Minimum Spanning Tree) algorithm, which are then filtered out to preserve the most relevant features out of each cluster. This increases the classification accuracy as well as running time and hence the computation cost. The efficacy of the presented approach was validated by comparing the same with other well-known feature selection algorithms: Fisher Score, CFS and ConsSF with respect to three classifiers: IB1, C4.5, and Naïve Bayes. The presented approach achieved higher Classification accuracy and lower run time.
16 May 2017 in Journal of Intelligent Information Systems
Sentiment analysis plays a very important role in BI's (Business Intelligence) applications which has been evident in the recent market activities. Towards sentiment analysis for most of the popular websites like Amazon, Facebook, Twitter necessitate the review of the customers which are used as a feedback. It's play very important role for product review, Business intelligence as well as in decision making. The main problem that arises to the point of view of users/customers is that, it is practically in-feasible to read all those online reviews one by one, because some of the products might have tens of thousand reviews. In this paper, reviews are collected from the sources like Amazon, Flipkart, and then used a method to combine both NLP (Natural Language Processing) and machine learning approach. Word sense disambiguation is also considered for this study. An improvised lesk algorithms is used for removing noise in the data. Different types of data have different types of properties and therefore are suited to different techniques correspondingly. This problem is closely related to the large scale nature of social networks and the necessity to perform aggregation operations, which results in the form of Pie-Chart. Thus, we aggregate millions of reviews into more user-friendly format.
1 July 2016 in International Journal of Service Science, Management, Engineering, and Technology
Now a day's application deal with Big Data has tremendously been used in the popular areas. To tackle with such kind of data various approaches have been developed by researchers in the last few decades. A recent investigated techniques to factored the data matrix through a known latent factor in a lower size space is the so called matrix factorization. In addition, one of the problems with the NMF approaches, its randomized valued could not provide absolute optimization in limited iteration, but having local optimization. Due to this, the authors have proposed a new approach that considers the initial values of the decomposition to tackle the issues of computationally expensive. They have devised an algorithm for initializing the values of the decomposed matrix based on the PSO. In this paper, the auhtors have intended a genetic algorithm based technique while incorporating the nonnegative matrix factorization. Through the experimental result, they will show the proposed method converse very fast in comparison to other low rank approximation like simple NMF multiplicative, and ACLS technique.
Now a days the enormity of High Dimensional data has been used in various real life applications. Most of the data mining techniques in descriptive analysis require a part of data sets into a fixed number of clusters based on user input, explicitly or learning by observation. For High Dimensional data set these fixed number of cluster given by user are not good estimation, because it leads to inefficient data distribution or it leads to various outlier. An efficient and scalable data mining technique is requires to deal with such type of data. In this paper we present a new algorithm to approach the problem of outlier detection in High Dimensional data with the help of descriptive analysis. Our technique is hybridization of density-based outlier detection and distance-based outlier detection technique.