Optimal feature extraction for bilingual OCR

Feature extraction in bilingual OCR is handicapped by the increase in the number of classes or characters to be handled. This is evident in the case of Indian languages whose alphabet set is large. It is expected that the complexity of the feature extraction process increases with the number of classes. Though the determination of the best set of features that could be used cannot be ascertained through any quantitative measures, the characteristics of the scripts can help decide on the feature extraction procedure. This paper describes a hierarchical feature extraction scheme for recognition of printed bilingual (Tamil and Roman) text. The scheme divides the combined alphabet set of both the scripts into subsets by the extraction of certain spatial and structural features. Three features viz geometric moments, DCT based features and Wavelet transform based features are extracted from the grouped symbols and a linear transformation is performed on them for the purpose of efficient representation in the feature space. The transformation is obtained by the maximization of certain criterion functions. Three techniques : Principal component analysis, maximization of Fisher's ratio and maximization of divergence measure have been employed to estimate the transformation matrix. It has been observed that the proposed hierarchical scheme allows for easier handling of the alphabets and there is an appreciable rise in the recognition accuracy as a result of the transformations.

Discriminative feature analysis and selection for document classification

Classification of a large document collection involves dealing with a huge feature space where each distinct word is a feature. In such an environment, classification is a costly task both in terms of running time and computing resources. Further it will not guarantee optimal results because it is likely to overfit by considering every feature for classification. In such a context, feature selection is inevitable. This work analyses the feature selection methods, explores the relations among them and attempts to find a minimal subset of features which are discriminative for document classification.

Feature selection for unsupervised learning

In this paper, we present a methodology for identifying best features from a large feature space. In high dimensional feature space nearest neighbor search is meaningless. In this feature space we see quality and performance issue with nearest neighbor search. Many data mining algorithms use nearest neighbor search. So instead of doing nearest neighbor search using all the features we need to select relevant features. We propose feature selection using Non-negative Matrix Factorization(NMF) and its application to nearest neighbor search. Recent clustering algorithm based on Locally Consistent Concept Factorization(LCCF) shows better quality of document clustering by using local geometrical and discriminating structure of the data. By using our feature selection method we have shown further improvement of performance in the clustering.

Cluster based training for scaling non-linear support vector machines

Support Vector Machines(SVMs) are hyperplane classifiers defined in a kernel induced feature space. The data size dependent training time complexity of SVMs usually prohibits its use in applications involving more than a few thousands of data points. In this paper we propose a novel kernel based incremental data clustering approach and its use for scaling Non-linear Support Vector Machines to handle large data sets. The clustering method introduced can find cluster abstractions of the training data in a kernel induced feature space. These cluster abstractions are then used for selective sampling based training of Support Vector Machines to reduce the training time without compromising the generalization performance. Experiments done with real world datasets show that this approach gives good generalization performance at reasonable computational expense.

Scalable non-linear Support Vector Machine using hierarchical clustering

This paper discusses a method for scaling SVM with Gaussian kernel function to handle large data sets by using a selective sampling strategy for the training set. It employs a scalable hierarchical clustering algorithm to construct cluster indexing structures of the training data in the kernel induced feature space. These are then used for selective sampling of the training data for SVM to impart scalability to the training process. Empirical studies made on real world data sets show that the proposed strategy performs well on large data sets.

A Unified Approach to Encoding and Classification Using Bimodal Projection-Based Features

In general the objective of accurately encoding the input data and the objective of extracting good features to facilitate classification are not consistent with each other. As a result, good encoding methods may not be effective mechanisms for classification. In this paper, an earlier proposed unsupervised feature extraction mechanism for pattern classification has been extended to obtain an invertible map. The method of bimodal projection-based features was inspired by the general class of methods called projection pursuit. The principle of projection pursuit concentrates on projections that discriminate between clusters and not faithful representations. The basic feature map obtained by the method of bimodal projections has been extended to overcome this. The extended feature map is an embedding of the input space in the feature space. As a result, the inverse map exists and hence the representation of the input space in the feature space is exact. This map can be naturally expressed as a feedforward neural network.

Decision tree design and applications in speech processing

The design and operation of the minimum cost classifier, where the total cost is the sum of the measurement cost and the classification cost, is computationally complex. Noting the difficulties associated with this approach, decision tree design directly from a set of labelled samples is proposed in this paper. The feature space is first partitioned to transform the problem to one of discrete features. The resulting problem is solved by a dynamic programming algorithm over an explicitly ordered state space of all outcomes of all feature subsets. The solution procedure is very general and is applicable to any minimum cost pattern classification problem in which each feature has a finite number of outcomes. These techniques are applied to (i) voiced, unvoiced, and silence classification of speech, and (ii) spoken vowel recognition. The resulting decision trees are operationally very efficient and yield attractive classification accuracies.

Sequential alternating proximal method for scalable sparse structural SVMs

Structural Support Vector Machines (SSVMs) have recently gained wide prominence in classifying structured and complex objects like parse-trees, image segments and Part-of-Speech (POS) tags. Typical learning algorithms used in training SSVMs result in model parameters which are vectors residing in a large-dimensional feature space. Such a high-dimensional model parameter vector contains many non-zero components which often lead to slow prediction and storage issues. Hence there is a need for sparse parameter vectors which contain a very small number of non-zero components. L1-regularizer and elastic net regularizer have been traditionally used to get sparse model parameters. Though L1-regularized structural SVMs have been studied in the past, the use of elastic net regularizer for structural SVMs has not been explored yet. In this work, we formulate the elastic net SSVM and propose a sequential alternating proximal algorithm to solve the dual formulation. We compare the proposed method with existing methods for L1-regularized Structural SVMs. Experiments on large-scale benchmark datasets show that the proposed dual elastic net SSVM trained using the sequential alternating proximal algorithm scales well and results in highly sparse model parameters while achieving a comparable generalization performance. Hence the proposed sequential alternating proximal algorithm is a competitive method to achieve sparse model parameters and a comparable generalization performance when elastic net regularized Structural SVMs are used on very large datasets.

Sparse coding based VLAD for efficient image retrieval

Representing images and videos in the form of compact codes has emerged as an important research interest in the vision community, in the context of web scale image/video search. Recently proposed Vector of Locally Aggregated Descriptors (VLAD), has been shown to outperform the existing retrieval techniques, while giving a desired compact representation. VLAD aggregates the local features of an image in the feature space. In this paper, we propose to represent the local features extracted from an image, as sparse codes over an over-complete dictionary, which is obtained by K-SVD based dictionary training algorithm. The proposed VLAD aggregates the residuals in the space of these sparse codes, to obtain a compact representation for the image. Experiments are performed over the `Holidays' database using SIFT features. The performance of the proposed method is compared with the original VLAD. The 4% increment in the mean average precision (mAP) indicates the better retrieval performance of the proposed sparse coding based VLAD.

Algebraic distributed space-time codes with low ML decoding complexity

"Extended Clifford algebras" are introduced as a means to obtain low ML decoding complexity space-time block codes. Using left regular matrix representations of two specific classes of extended Clifford algebras, two systematic algebraic constructions of full diversity Distributed Space-Time Codes (DSTCs) are provided for any power of two number of relays. The left regular matrix representation has been shown to naturally result in space-time codes meeting the additional constraints required for DSTCs. The DSTCs so constructed have the salient feature of reduced Maximum Likelihood (ML) decoding complexity. In particular, the ML decoding of these codes can be performed by applying the lattice decoder algorithm on a lattice of four times lesser dimension than what is required in general. Moreover these codes have a uniform distribution of power among the relays and in time, thus leading to a low Peak to Average Power Ratio at the relays.

On four-group ML decodable distributed space time codes for cooperative communication

A construction of a new family of distributed space time codes (DSTCs) having full diversity and low Maximum Likelihood (ML) decoding complexity is provided for the two phase based cooperative diversity protocols of Jing-Hassibi and the recently proposed Generalized Non-orthogonal Amplify and Forward (GNAF) protocol of Rajan et al. The salient feature of the proposed DSTCs is that they satisfy the extra constraints imposed by the protocols and are also four-group ML decodable which leads to significant reduction in ML decoding complexity compared to all existing DSTC constructions. Moreover these codes have uniform distribution of power among the relays as well as in time. Also, simulations results indicate that these codes perform better in comparison with the only known DSTC with the same rate and decoding complexity, namely the Coordinate Interleaved Orthogonal Design (CIOD). Furthermore, they perform very close to DSTCs from field extensions which have same rate but higher decoding complexity.

A Multilevel Inverter Scheme With Dodecagonal Voltage Space Vectors Based on Flying Capacitor Topology for Induction Motor Drives

This paper presents a multilevel inverter topology suitable for the generation of dodecagonal space vectors instead of hexagonal space vectors as in the case of conventional schemes. This feature eliminates all the 6n +/- 1 (n = odd) harmonics from the phase voltages and currents in the entire modulation range with an increase in the linear modulation range. The topology is realized by flying capacitor-based three-level inverters feeding from two ends of an open-end winding induction motor with asymmetric dc links. The flying capacitor voltages are tightly controlled throughout the modulation range using redundant switching states for any load power factor. A simple and fast carrier-based space-vector pulsewidth modulation (PWM) scheme is also proposed for the topology which utilizes only the sampled amplitudes of the reference wave for the PWM timing computation.

Feature Search in the Grassmanian in Online Reinforcement Learning

We consider the problem of finding the best features for value function approximation in reinforcement learning and develop an online algorithm to optimize the mean square Bellman error objective. For any given feature value, our algorithm performs gradient search in the parameter space via a residual gradient scheme and, on a slower timescale, also performs gradient search in the Grassman manifold of features. We present a proof of convergence of our algorithm. We show empirical results using our algorithm as well as a similar algorithm that uses temporal difference learning in place of the residual gradient scheme for the faster timescale updates.

Some remarks on the symplectic pairing on the moduli space of representations of the fundamental group of surfaces

This work grew out of an attempt to understand a conjectural remark made by Professor Kyoji Saito to the author about a possible link between the Fox-calculus description of the symplectic structure on the moduli space of representations of the fundamental group of surfaces into a Lie group and pairs of mutually dual sets of generators of the fundamental group. In fact in his paper [3] , Prof. Kyoji Saito gives an explicit description of the system of dual generators of the fundamental group.

An asymmetric homodinuclear complex containing the mu-x-oxobis(mu-x-carboxylato)diruthenium(III) core: X-ray structure of [Ru20(o2ec6H4-p-OMe)3(en) (PPh3)2] (C104)" 3CHC13

Asymmetric tri-bridged diruthenium(III) complexes, [Ru2O(O(2)CR)(3)(en) (PPh(3))(2)](ClO4) (R = C6H4-p-X: X = OMe (1a), Me (1b); en=1,2-diaminoethane), were prepared and structurally characterized. Complex 1a 3CHCl(3), crystallizes in the triclinic space group P (1) over bar with a = 14.029(5), b = 14.205(5), c = 20.610(6) Angstrom, alpha= 107.26(3), beta = 101.84(3), gamma= 97.57(3)degrees, V= 3756(2) Angstrom(3) and Z = 2. The complex has an {Ru-2(mu-O)(mu-O(2)CR)(2)(2+)} core and exhibits [O4PRu(mu-O)RuPO2N2](+) coordination environments for the metal centers. The novel structural feature is the asymmetric arrangement of ligands at the terminal sites of the core which shows an Ru... Ru separation of 3.226(3) Angstrom and an Ru-O-Ru angle of 119.2(5)degrees. An intense visible band observed near 570 nm is assigned to a charge transfer transition involving the d pi-Ru(III) and p pi-mu-O Orbitals. Cyclic voltammetry of the complexes displays a reversible Ru-2(III,III) reversible arrow Ru-2(III,IV) couple near 0.8 V (versus SCE) in MeCN-0.1 M TBAP.