Recognition of online handwritten music symbols

Paper submitted to MML 2013, 6th International Workshop on Machine Learning and Music, Prague, September 23, 2013.

Progressive structural analysis for dynamic recognition of on-line handwritten mathematical expressions

Structural analysis in handwritten mathematical expressions focuses on interpreting the recognized symbols using geometrical information such as relative sizes and positions of the symbols. Most existing approaches rely on hand-crafted grammar rules to identify semantic relationships among the recognized mathematical symbols. They could easily fail when writing errors occurred. Moreover, they assume the availability of the whole mathematical expression before being able to analyze the semantic information of the expression. To tackle these problems, we propose a progressive structural analysis (PSA) approach for dynamic recognition of handwritten mathematical expressions. The proposed PSA approach is able to provide analysis result immediately after each written input symbol. This has an advantage that users are able to detect any recognition errors immediately and correct only the mis-recognized symbols rather than the whole expression. Experiments conducted on 57 most commonly used mathematical expressions have shown that the PSA approach is able to achieve very good performance results.

Offline Handwriting Recognition Using Genetic Algorithm

In this paper, a new method for offline handwriting recognition is presented. A robust algorithm for handwriting segmentation has been described here with the help of which individual characters can be segmented from a word selected from a paragraph of handwritten text image which is given as input to the module. Then each of the segmented characters are converted into column vectors of 625 values that are later fed into the advanced neural network setup that has been designed in the form of text files. The networks has been designed with quadruple layered neural network with 625 input and 26 output neurons each corresponding to a character from a-z, the outputs of all the four networks is fed into the genetic algorithm which has been developed using the concepts of correlation, with the help of this the overall network is optimized with the help of genetic algorithm thus providing us with recognized outputs with great efficiency of 71%.

Interspecific variation in plumage colour among birds: species recognition or light environment?

The traditional explanation for interspecific plumage colour variation in birds is that colour differences between species are adaptations to minimize the risk of hybridization. Under this explanation, colour differences between closely related species of birds represent reproductive character displacement. An alternative explanation is that interspecific variation in plumage colour is an adaptive response to variation in light environments across habitats. Under this explanation, differences in colour between closely related species are a product of selection on signal efficiency. We use a comparative approach to examine these two hypotheses, testing the effects of sympatry and habitat use, respectively, on divergence in male plumage colour. Contrary to the prediction of the Species Isolation Hypothesis, we find no evidence that sympatric pairs of species are consistently more divergent in coloration than are allopatric pairs of species. However, in agreement with the Light Environment Hypothesis, we find significant associations between plumage coloration and habitat use. All of these results remain qualitatively unchanged irrespective of the statistical methodology used to compare reflectance spectra, the body regions used in the analyses, or the exclusion of areas of plumage not used in sexual displays. Our results suggest that, in general, interspecific variation in plumage colour among birds is more strongly influenced by the signalling environment than by the risk of hybridization.

Enhancing the immunogenicity and modulating the fine epitope recognition of antisera to a helical group A streptococcal peptide vaccine candidate from the M protein using lipid-core peptide technology

A conserved helical peptide vaccine candidate from the M protein of group A streptococci, p145, has been described. Minimal epitopes within p145 have been defined and an epitope recognized by protective antibodies, but not by autoreactive T cells, has been identified. When administered to mice, p145 has low immunogenicity. Many boosts of peptide are required to achieve a high antibody titre (> 12 800). To attempt to overcome this low immunogenicity, lipid-core peptide technology was employed. Lipid-core peptides (LCP) consist of an oligomeric polylysine core, with multiple copies of the peptide of choice, conjugated to a series of lipoamino acids, which acts as an anchor for the antigen. Seven different LCP constructs based on the p145 peptide sequence were synthesized (LCP1-->LCP7) and the immunogenicity of the compounds examined. The most immunogenic constructs contained the longest alkyl side-chains. The number of lipoamino acids in the constructs affected the immunogenicity and spacing between the alkyl side-chains increased immunogenicity. An increase in immunogenicity (enzyme-linked immunosorbent assay (ELISA) titres) of up to 100-fold was demonstrated using this technology and some constructs without adjuvant were more immunogenic than p145 administered with complete Freund's adjuvant (CFA). The fine specificity of the induced antibody response differed for the different constructs but one construct, LCP4, induced antibodies of identical fine specificity to those found in endemic human serum. Opsonic activity of LCP4 antisera was more than double that of p145 antisera. These data show the potential for LCP technology to both enhance immunogenicity of complex peptides and to focus the immune response towards or away from critical epitopes.

Genetic constraints on the evolution of mate recognition under natural selection

Field populations of Drosophila serrata display reproductive character displacement in cuticular hydrocarbons (CHCs) when sympatric with Drosophila birchii. We have previously shown that the naturally occurring pattern of reproductive character displacement can be experimentally replicated by exposing field allopatric populations of D. serrata to experimental sympatry with D. birchii. Here, we tested whether the repeated evolution of reproductive character displacement in natural and experimental populations was a consequence of genetic constraints on the evolution of CHCs. The genetic variance-covariance (G) matrices for CHCs were determined for populations of D. serrata that had evolved in either the presence or absence of D. birchii under field and experimental conditions. Natural selection on mate recognition under both field and experimental sympatric conditions increased the genetic variance in CHCs consistent with a response to selection based on rare alleles. A close association between G eigenstructure and the eigenstructure of the phenotypic divergence (D) matrix in natural and experimental populations suggested that G matrix eigenstructure may have determined the direction in which reproductive character displacement evolved during the reinforcement of mate recognition.

Pretraining Convolutional Neural Networks for Visual Recognition

Convolutional Neural Networks (CNN) have become the state-of-the-art methods on many large scale visual recognition tasks. For a lot of practical applications, CNN architectures have a restrictive requirement: A huge amount of labeled data are needed for training. The idea of generative pretraining is to obtain initial weights of the network by training the network in a completely unsupervised way and then fine-tune the weights for the task at hand using supervised learning. In this thesis, a general introduction to Deep Neural Networks and algorithms are given and these methods are applied to classification tasks of handwritten digits and natural images for developing unsupervised feature learning. The goal of this thesis is to find out if the effect of pretraining is damped by recent practical advances in optimization and regularization of CNN. The experimental results show that pretraining is still a substantial regularizer, however, not a necessary step in training Convolutional Neural Networks with rectified activations. On handwritten digits, the proposed pretraining model achieved a classification accuracy comparable to the state-of-the-art methods.

Doing Justice to Recognition

The traditional role of justice is to arbitrate where the good will of people is not enough, if even present, to settle a dispute between the concerned parties. It is a procedural approach that assumes a fractured relationship between those involved. Recognition, at first glance, would not seem to mirror these aspects of justice. Yet recognition is very much a subject of justice these days. The aim of this paper is to question the applicability of justice to the practice of recognition. The methodological orientation of this paper is a Kantian-style critique of the institution of justice, highlighting the limits of its reach and the dangers of overextension. The critique unfolds in the following three steps: 1) There is an immediate appeal to justice as a practice of recognition through its commitment to universality. This allure is shown to be deceptive in providing no prescription for the actual practice of this universality. 2) The interventionist character of justice is designed to address divided relationships. If recognition is only given expression through this channel, then we can only assume division as our starting ground. 3) The outcome of justice in respect to recognition is identification. This identification is left vulnerable to misrecognition itself, creating a cycle of injustice that demands recognition from anew. It seems to be well accepted that recognition is essentjustice, but less clear how to do justice to recognition. This paper is an effort in clarification.

Content Based Image Retrieval System for Malayalam Handwritten Characters

Content Based Image Retrieval is one of the prominent areas in Computer Vision and Image Processing. Recognition of handwritten characters has been a popular area of research for many years and still remains an open problem. The proposed system uses visual image queries for retrieving similar images from database of Malayalam handwritten characters. Local Binary Pattern (LBP) descriptors of the query images are extracted and those features are compared with the features of the images in database for retrieving desired characters. This system with local binary pattern gives excellent retrieval performance