Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 7 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
-------------------------------------------------------------------------------------------------------------------------------------------- 1203. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Textual,Topological and Graph Theoretic Face Recognition - Physique Recognition From Quadrilateral mesh isomorphism and Face Distinguisher from ImageNet predictions - related to 159,660,714,715,720,1185,1186,1187,1188,1189,1190,1201,1202 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search,Quadrilateral Mesh Deformation and Isomorphism,Maximum Common Subgraph Problem - 24 January 2022 -------------------------------------------------------------------------------------------------------------------------------------------- 1203.1 ImageNet Predictions from ResNet50 training dataset could filter visuals containing faces (human or animal) through textual predictions along with other objects in the frame which might be a preprocessing step in a Face recognition pipeline. 1203.2 First frame of an MP4 visual of 3 Ducks - testlogs/3Ducks_VID_20220119_173052.mp4Frame_1.jpg - is predicted by imagenet_videograph() and its ImageGraph computed by EventNet Tensor Products algorithm is written to logs testlogs/ImageGraph_Keras_Theano.log.24January2022 - following are vertices of the ImageGraph and their pageranks. Nature of objects and 3 animals in the visual are approximately predicted - including "face" as a vertex: [('wood_rabbit', 0.20490727007427248), ('meerkat', 0.1042849885504515), ('Bedlington_terrier', 0.06599733304073167), ('cliff', 0.04908043707463344), ('Wisconsinite', 0.044626573497110016), ('face', 0.023766317129790045), ('resident', 0.01613222743894782), ('Wisconsin', 0.01613222743894782), ('native', 0.01613222743894782), ('resemble', 0.014707510136039716), ('groomed', 0.014707510136039716), ('terrier', 0.014707510136039716), ('lamb', 0.014707510136039716), ('light', 0.014707510136039716), ('high', 0.013917572237708857), ('steep', 0.013917572237708857), ('rock', 0.013917572237708857), ('mongoose-like', 0.013337116985104445), ('South', 0.013337116985104445), ('viverrine', 0.013337116985104445), ('lemur', 0.013337116985104445), ('toes', 0.013337116985104445), ('like', 0.013337116985104445), ('Africa', 0.013337116985104445), ('four', 0.013337116985104445), ('common', 0.01316406639122259), ('North', 0.01316406639122259), ('brownish', 0.01316406639122259), ('rabbit', 0.01316406639122259), ('greyish', 0.01316406639122259), ('disease', 0.01316406639122259), ('tail', 0.01316406639122259), ('host', 0.01316406639122259), ('scapularis', 0.01316406639122259), ('America', 0.01316406639122259), ('fur', 0.01316406639122259), ('underside', 0.01316406639122259), ('pacificus', 0.01316406639122259), ('Ixodes', 0.01316406639122259), ('small', 0.01316406639122259), ('ticks', 0.01316406639122259), ('Lyme', 0.01316406639122259), ('white', 0.01316406639122259)] 1203.3 Textual predictions from imagenet_videograph() could be optionally added to decision tree in 1189 for improved Face recognition: if ImageNet Text Predictions contain face related vertices of high pagerank { /* Proceed with Topological and Graph Theoretic Face recognition */ } 1203.4 Full Physique quadrilateral mesh subgraph isomorphism algorithms try to match some fraction of 2 human bodies tessellated by quadrilateral mesh (only Delaunay triangulated mesh have been implemented thus far in NeuronRain) thus expanding the scope of Face Recognition to Physique Recognition, a comprehensive unique identification alternative to traditional face IDs - Figure 1.1,Figure 5.2 and Figure 5.3 of https://www.ics.uci.edu/~goodrich/pubs/approx-match.pdf approximately match two quadrilateral meshes topologically (two arbitrarily rigged animation characters) which could be adapted to two human visuals tessellated by quadrilateral mesh(s). 1203.5 Large Video File 3Ducks_VID_20220119_173052.mp4 for the frame earlier has been uploaded to https://kuja27.blogspot.com/2022/01/3-ducks-large-scale-visual-for.html
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1202 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1177 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1202. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Pairwise similarity loop and error handling - related to 159,660,714,715,720,1185,1186,1187,1188,1189,1190,1201 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search,Mesh Deformation - 21 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1202. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Pairwise similarity loop and error handling - related to 159,660,714,715,720,1185,1186,1187,1188,1189,1190,1201 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search,Mesh Deformation - 21 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1202.1 DeepLearning_ConvolutionNetwork_BackPropagation.py has been updated to loop through an array of images and compute pairwise face similarity by Graph Isomorphism (no subgraph isomorphism and iterations). Only full Isomorphism between Voronoi Facegraphs and Delaunay Trimesh(s) of a pair of images is being verified. 1202.2 8 face images of author are being matched for isomorphism spanning over a period of 19 years from 2003 to 2022. Images are of varied postures (passport,portrait-sitting,masked). New image testlogs/IMG_20220119_165911.jpg of the author wearing N95 COVID19 mask has been added to the computation to check if face isomorphism is recognized in disguise. 1202.3 Dlib shape detector fails to find 68 face landmarks for masked image testlogs/IMG_20220119_165911.jpg. Because of this an error condition is being handled to opt for contour centroids if 68 landmark centroids are not correctly found. Delaunay trimesh of testlogs/IMG_20220119_165911.jpg for contour centroids drawn by PyVis has been written to testlogs/IMG_20220119_165911_FaceRecognition_Segmentation_FaceGraph.html. PyVis drawn Delaunay Trimesh can be arbitrarily animated thus implementing a mesh deformation for arbitrary face expressions. PyVis mesh deformations can be non-planar. Voronoi Tessellation,Segmented Image and FaceGraph DOT files are written to: testlogs/IMG_20220119_165911-tessellated.jpg testlogs/IMG_20220119_165911_FaceRecognition_Segmentation_FaceGraph.dot testlogs/IMG_20220119_165911_segmented.jpg 1202.4 There are standard Mesh Deformation algorithms (As Rigid As Possible being one of them) for simulating the object animation and topological homeomorphic deformations in Computer Graphics. Identifying a face from its multiple mesh deformations is more accurate and harder than Graph Isomorphism recognition - facet areas change in addition to vertex renumbering. 1202.5 Logs for this update are written to testlogs/DeepLearning_ConvolutionNetwork_BackPropagation.log.21January2022. For 64 pairs of images (8*8), Full Graph Isomorphism is True for 19 pairs which is remarkable for a learner with no training priors. Subgraph isomorphism test has been avoided because of slowdown.
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Updated Requirements.txt
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1201 - year corrected
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1201. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Delaunay Triangulation Mesh Bezier Animation by PyVis - related to 159,660,714,715,720,1185,1186,1187,1188,1189,1190 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search - 17 January 2022,18 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1201. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Delaunay Triangulation Mesh Bezier Animation by PyVis - related to 159,660,714,715,720,1185,1186,1187,1188,1189,1190 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search - 17 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1201.1 DeepLearning_ConvolutionNetwork_BackPropagation.py has been updated to draw an animated HTML version of Delaunay Triangulation Mesh of a facial image by importing PyVis - https://pyvis.readthedocs.io/en/latest/tutorial.html#visualization. 1201.2 Delaunay Triangulation Mesh of a face aligns to 68 face landmarks and Bezier-like facets morphing of the Delaunay trimesh simulates the homeomorphic emotional deformations of the face (by changing facet areas) which is quintessential prerequisite for uniqueness of face irrespective of orientations. 1201.3 Delaunay Trimesh for facial image is written to an html file suffixed "_FaceRecognition_Segmentation_FaceGraph.html". An option to exit after full isomorphism check without subgraph isomorphism test and iterating the isomorphic subgraphs by VF2 and ISMAGS (both subgraph isomorphism test and iterations are time consuming for some photos) has been introduced by setting isomorphism_iterations to 0 in facegraph_similarity_metrics(). 1201.4 Two new images of the author - python-src/testlogs/Shrinivas_Kannan_01Sept2007_sitting.jpg (2007) and python-src/testlogs/SrinivasanKannan_alias_KaShrinivaasan_alias_ShrinivasKannan_photo.jpeg (2016) - have been added to pairwise graph isomorphism face similarity computation. 1201.5 Following are 6 isomorphic pairs of visuals (of 11 pairs of visuals of author) - from testlogs/DeepLearning_ConvolutionNetwork_BackPropagation.log.17January2022 which belong to 2016 and 2017 while remaining non-isomorphic 5 pairs have atleast one older image: ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071455.jpg', ' and ', 'testlogs/IMG_20160610_071603.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071455.jpg', ' and ', 'testlogs/IMG_20171112_180837.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071603.jpg', ' and ', 'testlogs/IMG_20171112_180837.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071455.jpg', ' and ', 'testlogs/SrinivasanKannan_alias_KaShrinivaasan_alias_ShrinivasKannan_photo.jpeg', ': 100.0') ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071603.jpg', ' and ', 'testlogs/SrinivasanKannan_alias_KaShrinivaasan_alias_ShrinivasKannan_photo.jpeg', ': 100.0') ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/SrinivasanKannan_alias_KaShrinivaasan_alias_ShrinivasKannan_photo.jpeg', ' and ', 'testlogs/IMG_20171112_180837.jpg', ': 100.0') 1201.6 11 pairs of visuals earlier span 14 years from 2003 to 2017 and belong to 2003,2007,2016,2017. As per decision tree in Section 1189,visuals of 2016 and 2017 (short durations) are heavily similar while isomorphism wanes between 2003 to 2007 (false negatives). Previous results reasonably establish graph-isomorphic face recognition of visuals separated by short durations. 1201.7 Approximate isomorphism test by faster_could_be_isomorphic() has been included in facegraph_similarity_metrics()
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1192 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1137 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1196. (THEORY and FEATURE) GIS Analytics - DBSCAN Density Clustering - Neural Network Thresholding implementation - related to 1151,1171,1174,1179,1194 and all sections on Merit of Large Scale Visuals, Text analytics, Unsupervised clustering, GIS and Urban Sprawl Analytics - Urban Extent Learners, Computational Geometry, Intrinsic Merit Rankings, R Ricker and SEDAC GPW4 Convex Hull Population Estimations, Community detection, Social Network Analytics, Space filling by Curves, Unequal Circle and Polynomial packing, Computational Geometry and Polygonal packing by Voronoi Tessellation, Four color theorem, Economic Merit, Computational Economics - GDP estimation from electricity consumption - 12 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1196. (THEORY and FEATURE) GIS Analytics - DBSCAN Density Clustering - Neural Network Thresholding implementation - related to 1151,1171,1174,1179,1194 and all sections on Merit of Large Scale Visuals, Text analytics, Unsupervised clustering, GIS and Urban Sprawl Analytics - Urban Extent Learners, Computational Geometry, Intrinsic Merit Rankings, R Ricker and SEDAC GPW4 Convex Hull Population Estimations, Community detection, Social Network Analytics, Space filling by Curves, Unequal Circle and Polynomial packing, Computational Geometry and Polygonal packing by Voronoi Tessellation, Four color theorem, Economic Merit, Computational Economics - GDP estimation from electricity consumption - 12 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1196.1 DBSCANClustering.py has been changed to implement a neural network thresholding of pixels in find_neighbours() and write_clustered_image() of DBSCAN algorithm. This feature has been made optional by a configurable parameter and original DBSCAN algorithm does not incorporate neural network thresholding in find_neighbours(). 1196.2 New clauses based on boolean flag "neuralnetwork" have been introduced in both find_neighbours() and write_clustered_image() which evaluate a neural network passed on as lambda function argument and filter the pixels above or below a threshold color thus detecting clusters or communities of urban areas. 1196.3 Function find_neighbours() computes neighbours satisfying the neural network while write_clustered_image() writes out the neural network clustered image to an image file suffixed "_neuralnetworkclustered.jpg". An option of choosing a square bounding box subset of an image has been provided in write_clustered_image() by "fraction" parameter which slices the image to subset square for invocation of find_neighbours() - epsilon neighbourhood is a percentage fraction of image dimension - so as to distinguish areas clustered and left unclustered by neural network lambda function. 1196.4 Global Human Settlement Layer (GHSL) image of Chennai Metropolitan area sprawl (of approximately 10000 square km by definition of metro sprawl in 1151 - semicircular sprawl covered by suburban train transport of 80 kms each in north,west and south) has been clustered by a simple neural network which weights each pixel by 2 for threshold value of 200. Clustered image has been written to GHSL_GIS_ChennaiMetropolitanArea_neuralnetworkclustered.jpg containing only densely populated localities excluding sparse suburbs and including composite city administrative area of 940 square km (Reference: GHSL convex hull and contour area estimator - R2019A - https://ghsl.jrc.ec.europa.eu/ucdb2018visual.php) and adjoining dense suburbs mostly concentrated in west,north and south. Lambda function argument can be arbitrary string and can depend on many variables not just limited to pixel value alone, computable as expression by eval(). 1196.5 Facebook High Resolution Settlement Layer (HRSL) Nighttime Streets imagery of world atlas - HRSL_World_NightTimeStreets.jpg - has been clustered by neural network thresholding - central 25% square fraction of image has been clustered and pixel colors of urban communities have been changed. Nighttime streets finetunes Nighttime lights urban area circles and is theoretically a graph of dotted streaks (edges) of streetlights in high resolution as against densely lit urban sprawl circles viewed from sky in VIIRS and MODIS. 1196.6 Four colors of Urban sprawl could be deciphered from a partition of Nighttime Street Graph as follows (For 29 color Urban GIS segmentation in 1174.8.5, percentage luminosity could be attributed accordingly): 1196.6.1 Manufacturing-IT-ITES - Vertices and Edges (Intersections and Streets) are extremely lit at night - highest electricity consumer 1196.6.2 Commercial - Vertices and Edges (Intersections and Streets) are highly lit at night - higher electricity consumer 1196.6.3 Residential - Vertices and Edges (Intersections and Streets) are less lit at night - high electricity consumer 1196.6.4 Greenery - Vertices and Edges (Intersections and Streets) are not lit almost (another metric: Greenery could be found by Vegetation index - NDVI for urban centres - https://ghsl.jrc.ec.europa.eu/ucdb2018visual.php) - minimal or no electricity consumed 1196.7 Neural network clustering is an independent feature (Continental US has been modelled by neural network extent learner for VIIRS and MODIS in 1175.12.8) which could be optionally used within DBSCAN find_neighbours(). Code invoking neural network clustering option in find_neighbours() has been commented within DBSCAN clustering().
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1258. (THEORY and FEATURE) Intrinsic Merit of Music - Kolmogorov Complexity approximation by Normalized Compression Distance (NCD) - implementation - 8 June 2022 - related to 110,1257 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Timeseries analysis, Majority Voting Vs Intrinsic - Condorcet Jury Theorem variants, Astronomical Datasets, Multiple Sequence Alignment of Strings, Handwriting Recognition --------------------------------------------------------------------------------------------------------------------------------------------- 1258.1 Intrinsic absolute Merit of music (independent of its enjoyability by majority voting) has been defined and implemented earlier by formal language gadgets - Minimum Description Length (Kraft Inequality) and Weighted Automata learnt from string of music waveform notes based on an axiomatic assumption that quality of music is proportional to fractal nature of notes string formalized by minimum size finite state automata accepting it and Kolmogorov complexity. 1258.2 In this commit,a new approximation of Kolmogorov complexity (or Normalized Information Distance which is Turing undecidable) - Normalized Compression Distance (NCD) - first introduced by [Cilibrasi-Vitanyi] as universal distance measure for datasets in "genomics, virology, languages, literature, music, handwritten digits, astronomy, and combinations of objects from completely different domains, using statistical, dictionary, and block sorting compressors", has been implemented as a function normalized_compression_distance() within MinimumDescLength.py. 1258.3 MinimumDescriptionLength.py and Entropy.py have been upgraded to Python 3.7. Usage of Zlib.compress() within MinimumDescLength.py has been now augmented with a bytes() binary cast and encoding to utf-8. 1258.4 Following are excerpts of logs - python-src/music_pattern_mining/testlogs/AudioToBitMatrix.log.KolmogorovComplexityNCD.8June2022 - which compute Minimum Descriptive Lengths (Kraft Inequality) and Normalized Compression Distance for 2 music waveforms of JS Bach: ('Merit of Audio - Minimum Descriptive Length and Entropy:', ('725.9117890720116', '1.9348455721183275')) ('Merit of Audio - Minimum Descriptive Length and Entropy:', ('733.1497901326836', '2.49102248187773')) Normalized Compression Distance (approximation of Kolmogorov Complexity) between 2 strings: 0.9917355371900827
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1195. (THEORY and FEATURE) Software Analytics - Degree Sequence Earth Mover Distance Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1193,1192,1191,581,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools,Graph Matching,Graph Similarity and Graph Isomorphism - 7 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1195. (THEORY and FEATURE) Software Analytics - Degree Sequence Earth Mover Distance Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1193,1192,1191,581,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools,Graph Matching,Graph Similarity and Graph Isomorphism - 7 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1195.1 python-src/software_analytics/CyclomaticComplexitySparkMapReducer.py - graph_matching() has been changed to implement Degree Sequence Earth Mover Distance similarity of two callgraphs. 1195.2 Degree Sequences of a callgraph are obtained from NetworkX degree() DegreeView and cast to dict() which is a map of vertices to their degrees. 1195.3 Degree Sequences are the reversed, sorted, monotone non-increasing values of the earlier dictionary of vertices to degrees. Degree Sequences of two callgraphs are matched by Wasserstein Earth Mover Distance similarity. 1195.4 Identical Degree Sequences of 2 graphs are less costly approximations of their isomorphism (Isomorphism => Identical Degree Sequences but Identical Degree Sequences !=> Isomorphism). 1195.5 3 pairs of callgraphs - 2 dissimilar and 1 identical - are matched, degree sequences are verified by is_graphical() and their approximate isomorphism is printed as boolean and Earth Mover Distance values. Identical callgraphs of same executable-parameter have 0 EMD which might differ depending on cybercrime penetration if the callgraphs are situated across multiple compromised machines which are vertices of SIS model infection graph (Refer: 619.3). Logs are written to testlogs/CyclomaticComplexitySparkMapReducer.log.DegreeSequenceEMD.7January2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1194 - References
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1194 - References
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1194 - References
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1194. (THEORY and FEATURE) GIS Analytics - DBSCAN Density Clustering implementation - related to 1182 and all sections on Merit of Large Scale Visuals, Text analytics, Unsupervised clustering, GIS and Urban Sprawl Analytics - Urban Extent Learners, Computational Geometry, Intrinsic Merit Rankings, R Ricker and SEDAC GPW4 Convex Hull Population Estimations, Community detection, Social Network Analytics, Space filling by Curves, Unequal Circle and Polynomial packing, Computational Geometry and Polygonal packing by Voronoi Tessellation, Four color theorem, Economic Merit, Computational Economics - 5 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1194. (THEORY and FEATURE) GIS Analytics - DBSCAN Density Clustering implementation - related to 1182 and all sections on Merit of Large Scale Visuals, Text analytics, Unsupervised clustering, GIS and Urban Sprawl Analytics - Urban Extent Learners, Computational Geometry, Intrinsic Merit Rankings, R Ricker and SEDAC GPW4 Convex Hull Population Estimations, Community detection, Social Network Analytics, Space filling by Curves, Unequal Circle and Polynomial packing, Computational Geometry and Polygonal packing by Voronoi Tessellation, Four color theorem, Economic Merit, Computational Economics - 5 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1194.1 New source file DBSCANClustering.py has been added to repository which implements award winning DBSCAN clustering algorithm especially useful for visual clustering of images, GIS Mapping and Urban Sprawl Extent Learners (NASA SEDAC,VIIRS and recent state-of-the-art Landsat9 TIRS2 and OLI2 - https://www.nasa.gov/content/landsat-9-image-gallery). 1194.2 DBSCAN is intuitive and finds clusters by density concentration of points around a core point which is the cluster centroid. 1194.3 Following is a gist of DBSCAN algorithm pseudocode implemented in DBSCANClustering.py: 1194.3.1 For each core point find the neighbours of it bounded by minPoints by some distance function 1194.3.2 If number of neighbours is < minPoints mark the point as noise else create a new cluster label for the core point 1194.3.3 Initialize a seed set to neighbours of the core point and for all neighbour points of a core point find the neighbours (neighbours of neighbours - expands the cluster recursively till seedset is exhausted) and add them to the cluster by labelling - Graph theoretically, community of neighbours or a strongly connected component of neighbourhood is detected. 1194.4 DBSCANClustering.py finds neighbours (greater than minpoints) within a 4*(epsilon)^2 pixel area square bounding box (of side 2*epsilon) around a point after filtering below a threshold pixel value. Clusters are stored in a dictionary map of pixel-to-label. Present implementation of DBSCAN is visual clustering specific but could be used for any multidimensional text dataset after minor datatype changes to find_neighbours() and clustering() functions. DBSCAN clusters could be optionally invoked in place of OpenCV2 Contouring, Watershed Segmentation and Minimum enclosing circle boundaries of urban sprawl extents in image_pattern_mining/ImageNet/GISUrbanSprawlAnalytics.py. 1194.5 An example scatterplot visual image_pattern_mining/ImageNet/testlogs/GHSL_GIS_ChennaiMetropolitanArea.jpg of Chennai Metropolitan Sprawl Population Density from Global Human Settlement Layer (GHSL - Urban Centre Database 2018 - https://ghsl.jrc.ec.europa.eu/ucdb2018visual.php# ) has been DBSCAN clustered and written to image_pattern_mining/ImageNet/testlogs/GHSL_GIS_ChennaiMetropolitanArea_dbscanclustered.jpg by write_clustered_image() - clusters are color coded by DBSCAN cluster label. Only the less dense areas of the map are clustered perhaps due to the thresholding. 1194.6 Class DBSCAN is parametrized by minimum number of neighbourhood points, epsilon size of the rectangular neighbourhood and threshold for choosing pixels. 1194.7 Logs are written to testlogs/DBSCANClustering.log.5January2022 which include number of clusters found and cluster labelled pixel dictionary.
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1192 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Updated Section 1192
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1193. (THEORY and FEATURE) Software Analytics - Approximate Graph Isomorphism Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1192,1191,581,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools - 2 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1193.1 CyclomaticComplexitySparkMapReducer.py - graph_matching() has been updated to invoke faster_could_be_isomorphic() for guesstimate of approximate isomorphism between two callgraphs or control flow graphs which is faster than fast_could_be_isomorphic() and exact but only presumptive. 1193.2 Approximate isomorphism has been made configurable by boolean flag approximateisomorphic. 1193.3 Two FTrace callgraph DOT files for pwd and find and two SATURN VIRGO linux kernel callgraph DOT files are matched for approximate isomorphism which turn out to be both False. 1193.4 Results are printed to logs testlogs/CyclomaticComplexitySparkMapReducer.log2.Python3.7.5.Isomorphism.2January2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1193. (THEORY and FEATURE) Software Analytics - Approximate Graph Isomorphism Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1192,1191,581,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools - 2 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1193.1 CyclomaticComplexitySparkMapReducer.py - graph_matching() has been updated to invoke faster_could_be_isomorphic() for guesstimate of approximate isomorphism between two callgraphs or control flow graphs which is faster than fast_could_be_isomorphic() and exact but only presumptive. 1193.2 Approximate isomorphism has been made configurable by boolean flag approximateisomorphic. 1193.3 Two FTrace callgraph DOT files for pwd and find and two SATURN VIRGO linux kernel callgraph DOT files are matched for approximate isomorphism which turn out to be both False. 1193.4 Results are printed to logs testlogs/CyclomaticComplexitySparkMapReducer.log2.Python3.7.5.Isomorphism.2January2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1192. (THEORY and FEATURE) Software Analytics - Graph Isomorphism Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1191,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools - 2 January 2022
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1192. (THEORY and FEATURE) Software Analytics - Graph Isomorphism Similarity of Code Callgraphs and Control Flow Graphs - Spark 3.0.1 + Python 3.7.5 - related to 1191,619,770,906,911,771 and all sections on Program Analysis, AsFer-USBmd-VIRGO-KingCobra FTrace kernel callgraph analysis - Valgrind-Callgrind-KCacheGrind-Flamegraph userspace callgraph analysis, SATURN control flow analysis of VIRGO linux kernel, Software Analytics, Pandemic model of Botnet-Defense adversarial Game theoretic security, Cybercrime analytics, Anti-Malware,Anti-Ransomware,Anti-Theft and Anti-Plagiarism tools - 2 January 2022 --------------------------------------------------------------------------------------------------------------------------------------------- 1192.1 Updated CyclomaticComplexitySparkMapReducer.py software analytics python implementation to define new function graph_matching() which finds Graph Isomorphisms between 2 control flow graphs or callgraphs of executables by VF2 and ISMAGS isomorphism implementations of NetworkX. 1192.2 Few pairs of SATURN VIRGO linux kernel DOT files, FTrace kernelspace callgraph DOT files, KCacheGrind userspace callgraph DOT files are compared by graph_matching() which internally instantiates NetworkX Graphs from DOT files. 1192.3 Isomorphisms (Full and Subgraph) are printed to logs testlogs/CyclomaticComplexitySparkMapReducer.log.Python3.7.5.Isomorphism.2January2022 and provide hints to kernelspace and userspace code most invoked by different program executables. Percentage similarities are printed in both directions of Graph matching (G1 -> G2 and G2 -> G1). Callgraphs are matched pairwise and only similar executables are fully isomorphic and subgraph isomorphic while dissimilar executables are not. Matching has been restricted to few pairs only as Graph isomorphism is computationally intensive and its lowerbound complexity is not yet proved. 1192.4 Graph Isomorphisms are tools to find similarity of codebases belonging to different repositories which is a pointer to frequently common algorithms and design patterns across implementations. Graph Isomorphism Detectors are used in anti-plagiarism tools and analyzing malicious code invocations (if Callgraphs-ControlFlowGraphs in two computers running similar executable are not isomorphic something could be brewing - one of them could be infected). 1192.5 Previous code change implements an item specified in https://github.com/shrinivaasanka/Krishna_iResearch_DoxygenDocs/blob/master/kuja27_website_mirrored/site/kuja27/PhDThesisProposal.pdf on isomorphism of Program Control Flow Graphs (SATURN-FTrace-Valgrind-Callgrind-KCacheGrind) and Slice Dependency Graphs. 1192.6 FTrace logs and DOT files of kernelspace callgraphs for "find" and "pwd" executables created by asfer_ftrace.sh and parsed to DOT by: ftrace_callgraph_dot("ftrace.pwd.log") ftrace_callgraph_dot("ftrace.find.log") have been added to repositories: find.ftrace_callgraph.dot ftrace.find.log ftrace.pwd.log pwd.ftrace_callgraph.dot 1192.7 Isomorphism iterations and ISMGAS symmetry have been made configurable arguments and set to default 25 and False respectively. CyclomaticComplexitySparkMapReducer.py has been upgraded to Python 3.7.5 + Spark 3.0.1 by autopep8 and 2to3 utilities. Python 2.7 version of CyclomaticComplexitySparkMapReducer.py has been renamed to CyclomaticComplexitySparkMapReducer.py.2.7.
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
1190. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Delaunay Tria ngulation Mesh isomorphism and Euler Characteristic - related to 159,660,714,715,720,1185,1186,1187,1188,1189 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Vi suals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgr aph),Graph matching,Graph similarity,Image Search - 30 December 2021
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 1190. (THEORY and FEATURE) People Analytics - Merit of Large Scale Visuals - Topological and Graph Theoretic Face Recognition - Delaunay Triangulation Mesh isomorphism and Euler Characteristic - related to 159,660,714,715,720,1185,1186,1187,1188,1189 and all sections on Topological Face and Handwriting recognition, Product Homotopy, Unique identification, Homeomorphism and Topological Deformations, Merit of Large Scale Visuals, People Analytics, Computational Geometry, Graph Edit Distance and Earth mover distance, Graph Isomorphism (Exact, Approximate and Subgraph),Graph matching,Graph similarity,Image Search - 30 December 2021 --------------------------------------------------------------------------------------------------------------------------------------------- 1190.1 DeepLearning_ConvolutionNetwork_BackPropagation.py has been changed to compute NetworkX Delaunay triangulation mesh graph dual of Voronoi Facegraph in draw_delaunay_triangulation(). Return values have been changed in some functions and new boolean clauses for Voronoi and Delaunay mesh(s) have been introduced. 1190.2 Euler characteristic of a facial image 2D Voronoi tessellation facegraph is printed to testlogs/DeepLearning_ConvolutionNetwork_BackPropagation.log.30December2021 (V-E+F) which is 2 for all images. This characteristic might vary for 3D Voronoi tessellation non-planar polyhedron mesh of a facial image: ('Euler characteristic of the Voronoi facegraph:', 2) ('Euler characteristic of the Voronoi facegraph:', 2) ('Euler characteristic of the Voronoi facegraph:', 2) ('Euler characteristic of the Voronoi facegraph:', 2) ('Euler characteristic of the Voronoi facegraph:', 2) 1190.3 Graph isomorphism of Delaunay triangulation mesh(s) for pairs of facial images is printed in testlogs/DeepLearning_ConvolutionNetwork_BackPropagation.log.30December2021 and contrary to intuition Delaunay triangulation dual mesh(s) which are localized to facial feature deformations are not isomorphic for none of the 6 pairs when Voronoi facegraphs are isomorphic for 3 pairs of face visuals separated by small time duration - Voronoi facegraphs may be best suited for face recognition by isomorphism: ------------------------- ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', True) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071455.jpg', ' and ', 'testlogs/IMG_20160610_071603.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True) ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', False) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True) ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', False) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True) ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', True) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071455.jpg', ' and ', 'testlogs/IMG_20171112_180837.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True) ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', True) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are isomorphic - percentage similarity - ', 'testlogs/IMG_20160610_071603.jpg', ' and ', 'testlogs/IMG_20171112_180837.jpg', ': 100.0') ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True) ('Voronoi FaceGraphs of two facial images are isomorphic - (True or False):', False) ('Delaunay Triangulation Mesh Graphs of two facial images are isomorphic - (True or False):', False) ('Voronoi FaceGraphs of two facial images are subgraph isomorphic - (True or False):', True)
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago
Section 1189 updated
Krishna iResearch - NeuronRain - K.Srinivasan
More... almost 5 years ago