Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 9 hours ago. based on code collected 1 day ago.
Nov 05, 2025 — Nov 05, 2026
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
815. (THEORY and FEATURE) Waveform Distance of Audios, Acoustic Distance of Strings, Name Similarity, Music Clustering - Python 3.7.5 - 27 March 2020 - related to 783 and all sections on People Analytics, Compressed Sensing, Intrinsic Merit, Syllable embedding of strings, Hyphenation and Music Pattern Mining
Srinivasan Kannan
More... over 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 815. (THEORY and FEATURE) Waveform Distance of Audios, Acoustic Distance of Strings, Name Similarity, Music Clustering - Python 3.7.5 - 27 March 2020 - related to 783 and all sections on People Analytics, Compressed Sensing, Intrinsic Merit, Syllable embedding of strings, Hyphenation and Music Pattern Mining --------------------------------------------------------------------------------------------------------------------------------------------- 1.NeuronRain AstroInfer already has a Music Recommender Systems implementation based on EMD similarities of Mel Frequency Cepstral Coefficients but not based on waveforms. 2.Usual Levenshtein Edit Distance is an alphabet based similarity measure while some problems in Name similarity require sophisticated acoustic matching algorithms (e.g NYSIIS, Soundex, Phonetic Match Rating) 3.This commit implements audio_distance() function for 2 audios (whether music or speech waveforms) which extracts waveform arrays of two audio(s) and compares them by Hausdorff distance and Earth Mover Distance which are more quantitative distance measures than edit distance. 4.Such a generic waveform phonetic distance between audios of two text strings is an alternative to Levenshtein Edit Distance. But it requires two human voice audio recordings which spells out the text of two strings (as in IVRs) 5.Syllables are language and script independent phonemes and previous audio_distance() by waveform match theoretically captures syllable phonetics better than edit distance and suits name similarity. Audio waveform match might perform better by spelling out Syllable hyphenated texts than plain texts. Syllable boundary word hyphenation has been implemented earlier for People Analytics and requires syllable to human voice audio transforms. 6.AudioToBitMatrix.py has been upgraded to Python 3.7.5. Logs for the change are at python-src/music_pattern_mining/testlogs/AudioToBitMatrix.log.AudioDistance.27March2020 which compare two similar and different pairs of audio clips of same duration.
Srinivasan Kannan
More... over 6 years ago
Theory - Zone Theorem and Ray shooting queries for approximate factors intersecting rasterized hyperbolic segment arrangement
Srinivasan Kannan
More... over 6 years ago
Theory - Factorization by Dynamic Parallel Planar Point Location
Srinivasan Kannan
More... over 6 years ago
814. (THEORY and FEATURE) Computational Geometric Factorization and Set Partition to Lagrange Four Square Theorem Square Tile Cover Reduction - Randomization - 26 March 2020 - related to 651, 728 and all sections on Set partitions, Shell Turing Machines, Kernel Lifting and Factorization
Srinivasan Kannan
More... over 6 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 814. (THEORY and FEATURE) Computational Geometric Factorization and Set Partition to Lagrange Four Square Theorem Square Tile Cover Reduction - Randomization - 26 March 2020 - related to 651, 728 and all sections on Set partitions, Shell Turing Machines, Kernel Lifting and Factorization -------------------------------------------------------------------------------------------------------------------------------------------- 1.DiscreteHyperbolicFactorizationUpperbound_TileSearch_Optimized.py has been changed to handle exceptions in Factor Spark Accumulator object by stripping off zero-es in zero() function. 2.Streaming_SetPartitionAnalytics.py has been changed for subprocess invocation commandline of factorization. Error handling for size of arrays has been done. 3.Hardcoded set has been removed and Set to be partitioned is instantiated by range() and maximum size of the set is sent by commandline sys.argv. Size of the set to partition is randomly chosen between 1 and maxsize and factorized to get the sides of the rectangle. Lagrangian Four Square Theorem Square Tile Cover reduction for the tile cover of factorized random set rectangle is printed. 4.This reduction has following phases: 4.1 Partition a set of random size in SymPy - random histogram 4.2 Factorize the size of the set of random size by Computational Geometry - sides of the rectangle 4.3 Reduce each part of the random partition to sum of four squares by Lagrange Four Square Theorem - kernel lifting from 1 dimensional partition to 2 dimensional rectangle partition. 4.4 Concatenate each of four square tiles per part of the random partition into single set of square tiles. 4.5 Tile the rectangle by square tiles. 4.6 Number of square tiles per tile cover of rectangle = 4 * number_of_parts_in_partition
Srinivasan Kannan
More... over 6 years ago
813. (FEATURE) Recursive Lambda Function Growth - corrections for machine translation - 25 March 2020^M
Srinivasan Kannan
More... over 6 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 813. (FEATURE) Recursive Lambda Function Growth - corrections for machine translation - 25 March 2020 -------------------------------------------------------------------------------------------------------------------------------------------- 1.RecursiveLambdaFunctionGrowth.py implementation has been changed to uncomment machine_translation(). But yet because of REST ratelimit English-Telegu translation is marred by "too many requests" exceptions. Exception handling and error checks have been added in machine_translation() for NoneType and ratelimit exceptions. Alternative could be local lookup of English-Telegu dictionary from flat file. 2.Self loop edges have been removed because of exceptions in rich club coefficient computation for the textgraph. 3.A news article on Chennai Urban Sprawl expansion has been mapped to textgraph and merit measures are printed to logs testlogs/RecursiveLambdaFunctionGrowth.log.MachineTranslation.25March2020. Telegu font decoding errors have created a Roman script graph. Proper noun filtering could further fine tune the objective meaning of the news article textgraph.
Srinivasan Kannan
More... over 6 years ago
Updated - game theory, chaos, machine learning models of epidemics and cybercrimes
Srinivasan Kannan
More... over 6 years ago
812. (THEORY and FEATURE) People Analytics - Human Resource Analytics - Recursive Lambda Function Growth Intrinsic Merit of Name-filtered Pro file - corrections, Chaotic Hidden Markov Model of Tenures/Attrition, Martingales - related to 418,420,613 - 25 March 2020
Srinivasan Kannan
More... over 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 812. (THEORY and FEATURE) People Analytics - Human Resource Analytics - Recursive Lambda Function Growth Intrinsic Merit of Name-filtered Profile - corrections, Chaotic Hidden Markov Model of Tenures/Attrition, Martingales - related to 418,420,613 - 25 March 2020 --------------------------------------------------------------------------------------------------------------------------------------------- 812.1.Recursive Lamda Function Growth Intrinsic Merit of a profile text has been benchmarked again on Python 3.7.5 upgraded implementation because of some misleading graph cores created in Python 2.7 by proper nouns (cities,organization names,institution names). Exceptions for word hyphenator maximum word length have been handled. Logs are at python-src/testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.25March2020 812.2.An example profile text (testlogs/CV.tex) of the author has been analyzed and dense subgraph analysis of its textgraph has been executed. Most of the proper nouns have been filtered from profile text but for few ambiguities (e.g "Sun","Global" are in dictionary but they were old first names of organizations author worked for and "Green" is name of NeuronRain repositories, "Gold medalist" is mistook for "Golf medalist") 812.3.By and large top percentile core numbers of the profile textgraph reasonably reflect the stature of the profile and name filtering increases relevance of the dense subgraphs. Though anything can be written on a CV, an id verified profile content is a trustworthy measure of merit (Trusted Social Network Connections of professional profile in organizations worked vouch for authenticity despite being fame based citations) and facilitates autonomous recruitment if Merit Lowerbounds Fame: ============================================================================================================= Unsupervised Classification based on top percentile Core numbers of the definition graph(subgraph of WordNet) ============================================================================================================= ('This document belongs to class:', 'senior', ',core number=', 4) ('This document belongs to class:', 'architect', ',core number=', 4) ('This document belongs to class:', 'college', ',core number=', 4) ... ('This document belongs to class:', 'group', ',core number=', 2) ('This document belongs to class:', 'swarm', ',core number=', 2) ('This document belongs to class:', 'belongs', ',core number=', 2) ('This document belongs to class:', 'Green', ',core number=', 2) ('This document belongs to class:', 'golf', ',core number=', 2) ('This document belongs to class:', 'medalist', ',core number=', 2) ('This document belongs to class:', 'essential', ',core number=', 2) ('This document belongs to class:', 'kernel', ',core number=', 2) ... ('This document belongs to class:', 'technical_school', ',core number=', 2) ('This document belongs to class:', 'semantic', ',core number=', 2) ... ('This document belongs to class:', 'expert', ',core number=', 2) ('This document belongs to class:', 'adviser', ',core number=', 2)
Srinivasan Kannan
More... over 6 years ago
Updated documentation on mining celestial data - PrefixSpan
Srinivasan Kannan
More... over 6 years ago
Chaotic PRG - Econometrics - Verhulst-Pearl-Reed Logistic Law
Srinivasan Kannan
More... over 6 years ago
Chaotic PRG implementation - modelling econometric,biological,social bigdata sets by Logistic equation
Srinivasan Kannan
More... over 6 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 6 years ago
811. (THEORY and FEATURE) Interview Algorithm - Spark Recursive Gloss Overlap Implementation - Python 3.7.5 upgrade and Quadcore Benchmarks - 20 March 2020
Srinivasan Kannan
More... over 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 811. (THEORY and FEATURE) Interview Algorithm - Spark Recursive Gloss Overlap Implementation - Python 3.7.5 upgrade and Quadcore Benchmarks - 20 March 2020 --------------------------------------------------------------------------------------------------------------------------------------------- 1.Spark Recursive Gloss Overlap MapReduce implementation in python-src/InterviewAlgorithm has been upgraded to Python 3.7.5 and benchmarked on Quadcore. 2.Earlier implementation on Python 2.7 and Dual core was slower because of local[2] which has been upgraded to local[4] for quadcores. 3.Memcache python client has been changed to pymemcache for Python 3.7.5 and pymemcache API changes have been incorporated. 4.Spark Python Recursive Gloss Overlap sources have been compiled to C by Cython3 (upgraded from Cython2) - following are commandline alternatives to setup.py and .so(s): cython3 <Python3.7.5_file.py> -o <Cython3_file.c> gcc -I<include_Python3.7_headers> <Cython3_file.c> -o <Cython3_executable> -lpython3.7m 5.Unicode encode() for Python 3 strings has been invoked wherever necessary 6.Pickle file is opened in 'a' and not 'ab' because of Python 3 bytes type conflict 7.Python 2.7 versions still depend on memcache python client, have .2.7 extension and Cython 2 .so(s) remain intact. 8.Replacement of Flat file pickle in MapReduce by memcache get/set was tried but the code for it has been left commented because of huge number of connection reset errors within memcache. This is despite lockdown of pages in memcache.conf and parallel connections set to 1024. In-memoy cache of Synsets in Spark mapreduce could reduce latency further and might be uncommented later. 9.2 sentences in small one liner file have been mapped to 2 textgraphs and timestamped textgraph plot images and logs are at InterviewAlgorithm/testlogs/
Srinivasan Kannan
More... over 6 years ago
810. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Delaunay Triangulation and Voronoi diagram of Segmented Image, Intrinsic Merit of Large Scale Visuals, Computational Geometry - Planar Point Location, Computational Geometric Factorization - Updated
Srinivasan Kannan
More... over 6 years ago
Face Graphs of Urban Sprawls - Fair Multi Agent Graph Coloring
Srinivasan Kannan
More... over 6 years ago
810. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Delaunay Triangulation and Voronoi diagram of Segmented Image, Intrinsic Merit of Large Scale Visuals, Computational Geometry - Planar Point Location, Computational Geometric Factorization - 16 March 2020, 17 March 2020
Srinivasan Kannan
More... over 6 years ago
810. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Delaunay Triangulation and Voronoi diagram of Segmented Image - 16 March 2020
Srinivasan Kannan
More... over 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 810. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Delaunay Triangulation and Voronoi diagram of Segmented Image - 16 March 2020 --------------------------------------------------------------------------------------------------------------------------------------------- 1.ImageGraph_Keras_Theano.py image_segmentation() has been updated to approximately compute face graph of a segmented image by Voronoi diagram for segment centroids. 2.Centroids from connected components statistics are construed to be the points of a Voronoi diagram planar subdivision. 3.A planar rectangle of the size of image is instantiated by Subdiv2D() and centroids of image segment components are populated in planar subdivision. 4.Delaunay triangles and Voronoi faces tesselation of the planar subdivision are obtained by OpenCV API - getTriangleList() and getVoronoiFacetsList(). 5.Voronoi diagram draws boundaries optimally in a way such that centroids are equidistant from them and thus approximates the face graph for the connected segment components of the image which is essentially necessary to simulate envy-free allocation of four categories/colors of an urban sprawl and a fair division MARA. An alternative to Voronoi diagram could be to minimize a linear program of average weighted distance between centroids of segment faces - optimal weights of edges between face centroids decide the dimensions of every segment category. 6.For every facet in the Voronoi diagram which corresponds to a segment, a NetworkX vertex object is created by stringified concatentation of pixel ordinates delimited by "#".This uniquely identifies a vertex in facegraph. Edges are added for each face boundary in NetworkX Undirected graph. This NetworkX facegraph of an image is sufficient for graph theoretic analysis of any image. 7.Polygon is drawn for each face and NetworkX facegraph of an image in its entirety is written to a DOT file. 8.Two GIS images of Chennai Urban Sprawl have been segmented by watershed and their facegraphs have been plotted as PNG images. (There are graph theoretic alternatives to watershed viz., GRABCUT graph based image segmentation algorithm which is on similar line) 9.New video montage - Google Search of NeuronRain repositories and Google Scholar profile of author https://scholar.google.com/citations?user=eLZY7CIAAAAJ&hl=en - has been recorded as MP4 and analyzed for EventNet Tensor Product Merit which is a good mix of visuals and text. Incidentally every frame of a video can also be segmented graph theoretically as FaceGraph as against ImageGraph analysis from ImageNet predictions. 10. Segment centroid Voronoi diagrams are as well irregular bounding boxes implementations for an image as opposed to rectangles.
Srinivasan Kannan
More... over 6 years ago
793. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Image Segmentation and Face components statistics,Dual Graph,Four color theorem
Srinivasan Kannan
More... over 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 793. (THEORY and FEATURE) GIS and Urban Sprawl Analytics - Image Segmentation and Face components statistics,Dual Graph,Four color theorem - 12 March 2020 --------------------------------------------------------------------------------------------------------------------------------------------- 1.ImageNet Keras-Theano + OpenCV + Python 3.7.5 implementation - ImageGraph_Keras_Theano.py - has been changed for computing statistics of image segmentation by watershed algorithm. 2.image_segmentation() function has been augmented to invoke cv2.connectedComponentsWithStatsWithAlgorithm(sure_fg,connectivity=8,ltype=2,ccltype=cv2.CCL_GRANA) which computes the connected face components of foreground image accompanied by statistics and algorithm parameters. 8-way connectivity of an image by GRANA connected components algorithm is computed. cv2.connectedComponents() is limited to markers only. 3.Labels,Centroids and Statistics of each connected component of a segmented image are printed. Every component of segmented image is also a face in graph theoretic terms and a dual graph of adjacent faces can be drawn connecting centroids of faces. 4.By Four color theorem (a kind of Multiple Agent Resource Allocation), centroids of adjacent faces could be made color-repulsive - every face and thus its centroid can be colored by one of the four categories (Manufacturing-IT-ITES,Residential,Commercial,Greenery) and adjacent faces avoid same color. Dense subgraphs of Colored Weighted Dual graph (vertices are colored by one of the 4 categories and edges are labelled by distance between centroids of the faces) of segmented urban sprawl GIS offer insight into urbanization, expansion and density. 5.An example montage MP4 video (Example Video 1 - Erstwhile Research Scholar profile - 1995-2011 - academics and industry - of the author) has been benchmarked again along with a segmentation of Chennai Urban Sprawl GIS JPEG (SEDAC - http://sedac.ciesin.columbia.edu/mapping/popest/gpw-v4/). 6.Logs at testlogs/ImageGraph_Keras_Theano.log.12March2020 show two prominent segments of Chennai Urban Sprawl and their centroids.
Srinivasan Kannan
More... over 6 years ago
Theory - Computational Geometric Factorization - Rasterization of Hyperbolic arc by chain of line segments - Bresenham algorithm - Graphics perspective
Srinivasan Kannan
More... over 6 years ago
Theory - Generalized Condorcet Jury Theorem for Multipartisan Elections, Theoretical EVMs, Set Partitions - Reference
Srinivasan Kannan
More... over 6 years ago
Computational Geometric Factorization - Fast Queries in Planar Point Location - Reference
Srinivasan Kannan
More... over 6 years ago
Computational Geometric Factorization - Entropy based Planar Point Location - Reference
Srinivasan Kannan
More... over 6 years ago
Updated caution on commercial deployments
Srinivasan Kannan
More... over 6 years ago
(THEORY and FEATURE) Computational Geometric Factorization - References on Rasterization and Point Location, practical alternatives to PRAM - BSP,BSP*,LogP,CGM
Srinivasan Kannan
More... over 6 years ago