Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 8 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
1127. (THEORY and FEATURE) People Analytics - HR and Talent Analytics - Domain Specific Dictionary - LinkedIn Dataset analytics in Spark 3.0.1+ Hadoop 3.2 - related to 613,727,812,876 and all sections on People Analytics, Intrinsic Merit Versus Fame Equilibrium, Novelty Detection, BKS Conjecture, Stability of Interview LTF/PTF - 19 December 2020
Srinivasan Kannan
More... almost 6 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 1127. (THEORY and FEATURE) People Analytics - HR and Talent Analytics - Domain Specific Dictionary - LinkedIn Dataset analytics in Spark 3.0.1+ Hadoop 3.2 - related 613,727,812,876 and all sections on People Analytics, Intrinsic Merit Versus Fame Equilibrium, Novelty Detection, BKS Conjecture, Stability of Interview LTF/PTF - 19 December 2020 -------------------------------------------------------------------------------------------------------------------------------------------- 1.NeuronRain People Analytics is mainly about automating the process of recruiting talent by minimizing or obviating human error and intervention. 2.Theoretically if intrinsic merit lowerbounds fame, automatic recruitment is feasible which has been analyzed in earlier relevant sections. In this commit, profile parser function has been updated to include a domain specific dictionary argument which contains fields specific to a work domain - IT, Manufacturing, Academics etc., 3.For Information Technology domain, example dictionary has two fields (values) - "name" (Information Technology) and "opensource_sloc" (Number of Source Lines of Code contributed to OpenSource products). Domain of IT is quite different from other industries in the sense work experience and technical wherewithal is quantifiable by lines of code and arbitrary number of clones can be derived from digital content which is impossible in other disciplines. 4.Technical Talent in Information Technology domain is measurable quantitatively by number of lines of code written which is used in various opensource codebase analyzers (e.g OpenHub COCOMO analyzer) and software complexity metrics (e.g Function Point Analysis). 5.Total experience of a profile is weighted by opensource_sloc value (which is at present just a linear addition function but could be any other) for least energy intrinsic merit of a profile which is the single largest crucial parameter more important than mere tenure duration for automatic recruitment in Information Technology domain and more intrinsic than Majority Voting (which suffers from limitations of Condorcet Jury Theorem) and manual processes of Interview (which is subject to BKS Conjecture and human error). 6.If Novelty Detection measures (e.g Patents) and SLOC measures per proprietary and opensource products of organizations are accumulated and weighted it could further augment the previous merit statistic. 7.Spark Version has been upgraded to 3.0.1 (based on Hadoop 3.2) for LinkedIn Dataset Analytics.
Srinivasan Kannan
More... almost 6 years ago
1126. (THEORY and FEATURE) People Analytics - Word Sense Disambiguation changed to NeuronRain implementation - related to 876,1123 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics, Set Partition and Histogram analytics - 16 December 2020
Srinivasan Kannan
More... almost 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 1126. (THEORY and FEATURE) People Analytics - Word Sense Disambiguation changed to NeuronRain implementation - related to 876,1123 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics, Set Partition and Histogram analytics - 16 December 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1. Recursive Lambda Function Growth Meaning Representation was inferring some misleading lambda functions related to the proper noun dictionary-filtered profile contents - Word "Architect" was incorrectly disambiguated to civil engineering and produced lemmas - "slab", "home_plate", "base_ball" et al. 2. This was perhaps because of deficiencies of NLTK lesk implementation. Changing to PyWSD significantly slows down and there are WordNet errors within PyWSD. This has necessitated reverting to NeuronRain Lesk Word Sense Disambiguation function best_matching_synset() and both NLTK and PyWSD disambiguation have been disabled. 3. NeuronRain WSD best_matching_synset() has extracted quite accurate synset lemmas which have been used to create TextGraph and Graph Tensor Neural Network Lambda functions are grown from Random walks and Cycles on this new TextGraph - TextGraph for best_matching_synset() is much different from those of NLTK/PyWSD Lesk functions. 4. New TextGraph has been written to DOT file RecursiveLambdaFunctionGrowth.dot 5. Logs testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.16December2020 print following Dense subgraph classes (which give high weightage to durations) and intrinsic merit parameters while Betweenness,Closeness,Degree,PageRank centralities rank the profile highly in class "Software": ... ============================================================================================================= Unsupervised Classification based on top percentile Core numbers of the definition graph(subgraph of WordNet) ============================================================================================================= ('This document belongs to class:', 'April', ',core number=', 6) ('This document belongs to class:', 'March', ',core number=', 6) ('This document belongs to class:', 'May', ',core number=', 6) ('This document belongs to class:', 'month', ',core number=', 6) ('This document belongs to class:', 'preceding', ',core number=', 6) ('This document belongs to class:', 'following', ',core number=', 6) ('This document belongs to class:', 'architect', ',core number=', 4) ('This document belongs to class:', 'search', ',core number=', 4) ('This document belongs to class:', 'Hindi', ',core number=', 4) ('This document belongs to class:', 'institute', ',core number=', 4) ('This document belongs to class:', 'education', ',core number=', 4) ('This document belongs to class:', 'experience', ',core number=', 4) ('This document belongs to class:', 'software', ',core number=', 4) ('This document belongs to class:', 'science', ',core number=', 4) ('This document belongs to class:', 'activities', ',core number=', 4) ('This document belongs to class:', 'Tamil', ',core number=', 4) ('This document belongs to class:', 'skill', ',core number=', 4) ('This document belongs to class:', 'knowledge', ',core number=', 4) ('This document belongs to class:', 'something', ',core number=', 4) ('This document belongs to class:', 'India', ',core number=', 4) ('This document belongs to class:', 'written', ',core number=', 4) ('This document belongs to class:', 'language', ',core number=', 4) ('This document belongs to class:', 'spoken', ',core number=', 4) ('This document belongs to class:', 'someone', ',core number=', 4) ... intrinsic_merit_dict: {'graph_tensor_neuron_network_intrinsic_merit': 1011.984847256428, 'maximum_per_random_walk_graph_tensor_neuron_network_intrinsic_merit': ('(experience(content,(education(skill,(cognition(ability,(cognition(ability,(psychological_feature(cognition,(activity(event,(act(education,(skill(education,(act(activity,(psychological_feature(institute,(abstraction(event,(organization(group,(social_group(institute,(organization(association,(abstraction(association,(group(science,(abstraction(knowledge_domain,(cognition(abstraction,(psychological_feature(content,(discipline(social_group,psychological_feature))))))))))))))))))))))))))))))))))))))))', 16.89555722055722), 'korner_entropy': 0.7184074850146093, 'density': 0.007855421686746989, 'bose_einstein_intrinsic_fitness': 0.8612343424751857, 'recursive_gloss_overlap_intrinsic_merit': 856492032.0, 'empath_sentiment': (0.0, 0.0, 1.8548942433184686e-296)}
Srinivasan Kannan
More... almost 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 1125. (THEORY and FEATURE) Software Analytics - Cyclomatic Complexity Program Analyzer - Graph Edit Distance - related to 581,1124 and all sections on Program Analysis, Software Analytics, Graph Edit Distance, Inexact Graph Matching - 15 December 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1. Cyclomatic Complexity Program Analyzer Spark implementation python-src/software_analytics/CyclomaticComplexitySparkMapReducer.py has been refactored and new function cyclomatic_complexity() has been defined for Cyclomatic Analysis of FTrace callgraph by GraphFrames. 2. Function graph_mining() now invokes GraphEditDistance() from GraphMining_GSpan for each pair of callgraph NetworkX objects instantiated from DOT files in addition to GSPAN algorithm for frequent subgraph mining. 3. Finding code similarity by Graph Edit Distance between callgraphs is dynamic program analysis (Inexact matching) which is more comprehensive profiling than Graph Isomorphism (Exact Matching)
Srinivasan Kannan
More... almost 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 1125. (THEORY and FEATURE) Software Analytics - Cyclomatic Complexity Program Analyzer - Graph Edit Distance - related to 581,1124 and all sections on Program Analysis, Software Analytics, Graph Edit Distance, Inexact Graph Matching - 15 December 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1. Cyclomatic Complexity Program Analyzer Spark implementation python-src/software_analytics/CyclomaticComplexitySparkMapReducer.py has been refactored and new function cyclomatic_complexity() has been defined for Cyclomatic Analysis of FTrace callgraph by GraphFrames. 2. Function graph_mining() now invokes GraphEditDistance() from GraphMining_GSpan for each pair of callgraph NetworkX objects instantiated from DOT files in addition to GSPAN algorithm for frequent subgraph mining. 3. Finding code similarity by Graph Edit Distance between callgraphs is dynamic program analysis (Inexact matching) which is more comprehensive profiling than Graph Isomorphism (Exact Matching)
Srinivasan Kannan
More... almost 6 years ago
----------------------------------------------------------------------------------------------------------------------------------------- 1124. (THEORY and FEATURE) Graph Search and Analytics - NetworkX Graph Edit Distance - related to 753, 829, 1123 and all sections on TextGraph algorithms for Meaning Representation, Social Network Analytics, Distance and Similarity measures for Text/Audio/Video/People, Frequent Subgraph mining, Inexact Graph Matching, Program Analysis - 9 December 2020 -----------------------------------------------------------------------------------------------------------------------------------------
Srinivasan Kannan
More... almost 6 years ago
----------------------------------------------------------------------------------------------------------------------------------------- 1124. (THEORY and FEATURE) Graph Search and Analytics - NetworkX Graph Edit Distance - related to 753, 829, 1123 and all sections on TextGraph algorithms for Meaning Representation, Social Network Analytics, Distance and Similarity measures for Text/Audio/Video/People, Frequent Subgraph mining, Inexact Graph Matching, Program Analysis - 9 December 2020 ----------------------------------------------------------------------------------------------------------------------------------------- 1.GraphMining_GSpan Python implementation has been upgraded to Python 3.7 by autopep8 and 2to3-3.7 utilities. 2.New function GraphEditDistance() has been defined in GSpan class which takes two NetworkX Graph objects as arguments and computes exact Graph Edit Distance (NP-Hard) by graph_edit_distance() and approximations by optimize_edit_paths() and optimize_graph_edit_distance() NetworkX APIs. 3.Logs in python-src/testlogs/GraphMining_GSpan.GraphEditDistance.log.9December2020 capture the GSpan Graph mining for some example TextGraphs (Recursive Gloss Overlap algorithm) and Graph Edit Distance for two example numeric labelled graphs. 4.GraphEditDistance() could be invoked from any Graph analytics implementation in NeuronRain for Similar Graph clustering viz., Social network graphs, FaceGraphs (Delaunay triangles and Voronoi facets) of Segmented images(Pattern recognition, GIS, Urban Sprawls), TextGraphs, Economic networks, Program Slice Graphs, FTrace kernel callgraphs, Userspace Valgrind/KCacheGrind/Callgrind callgraphs, Control Flow Graphs et al. 5.Graph Edit Distance for Audios (Speech) assumes a hypothetical AudioNet similar to ImageNet dataset or Speech-to-Text engine for eliciting graph structure in speech though Music as a graph is vague to define, but NeuronRain music pattern mining learns a weighted automata from musical clips which are directed graphs of state transitions labelled by heptatonic western classical notes (CDEFGABC). 6.Classical Indian music has 72 fundamental notes patterns and many derivatives from them.
Srinivasan Kannan
More... almost 6 years ago
1123. (THEORY and FEATURE) People Analytics - Alphabet Vectorspace Embedding - Earth Mover Distance between String Syllable Tensors - related to 752,874,875 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics, Set Partition and Histogram analytics - 27 November 2020
Srinivasan Kannan
More... almost 6 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 1123. (THEORY and FEATURE) People Analytics - Alphabet Vectorspace Embedding - Earth Mover Distance between String Syllable Tensors - related to 752,874,875 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics, Set Partition and Histogram analytics - 27 November 2020 -------------------------------------------------------------------------------------------------------------------------------------------- 1.This commit augments CompressedSensing.py - alphabet_vectospace_embedding_distance() function to optionally compute syllable vector earth mover distance between two phonetic syllable hyphenated string 2D tensors (encoded by ordinal integers) by syllable_tensor_distance boolean flag 2.Lot of options were tried for computing distance norm between two 2D tensors having arbitrary number of rows and columns of unequal lengths - TensorFlow, sklearn, SciPy directed_hausdorff, wasserstein_distance and cdist, OpenCV2 EMD - and OpenCV2 EMD has been chosen because of its implicit dimension expansion, support for unequal rows and columns between two 2D tensors (which is very much necessary for comparing two strings of unequal number of syllables of arbitray length) and ability to measure syllables as histograms - syllables partition a string to histogram buckets and distance between two strings is the amount of work necessary to move one string syllable histogram to another (2D analogue of Edit Distance) which is captured by Earth Mover Distance of OpenCV. 3.Function tosig() translates a syllable hyphenation of string to OpenCV.EMD() format. 4.Logs at python-src/testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.27November2020 compute EMD between following string syllable hyphenations: csensing.alphabet_vectorspace_embedding_distance([['n','o'], ['m','i'],['n','a','l']], [['s','e'], ['m','i'],['n','a','l']],True) csensing.alphabet_vectorspace_embedding_distance([['t','e','r'], ['m','i'],['n','a','l']], [['s','e'], ['m','i'],['n','a','l']],True) csensing.alphabet_vectorspace_embedding_distance([['S','h','r','i'],['n','i'],['v','a','a'],['s','a','n']],[['S','r','i'],['n','i'],['v','a','s','a','n']],True) 5.Following are the respective EMDs computed for string syllable hyphenations(encoded by ordinals): ordembedding1: [[110, 111], [109, 105], [110, 97, 108]] ordembedding2: [[115, 101], [109, 105], [110, 97, 108]] Earth Mover Distance between two String syllable tensors: (0.006666666828095913, None, array([[110., 0., 0., 0., 0., 0., 0.], [ 5., 101., 0., 0., 0., 0., 0.], [ 0., 0., 109., 0., 0., 0., 0.], [ 0., 0., 0., 105., 0., 0., 0.], [ 0., 0., 0., 0., 110., 0., 0.], [ 0., 0., 0., 0., 0., 97., 0.], [ 0., 0., 0., 0., 0., 0., 108.]], dtype=float32)) ordembedding1: [[116, 101, 114], [109, 105], [110, 97, 108]] ordembedding2: [[115, 101], [109, 105], [110, 97, 108]] Earth Mover Distance between two String syllable tensors: (0.0, None, array([[115., 0., 0., 0., 0., 0., 0.], [ 0., 101., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 109., 0., 0., 0., 0.], [ 0., 0., 0., 105., 0., 0., 0.], [ 0., 0., 0., 0., 110., 0., 0.], [ 0., 0., 0., 0., 0., 97., 0.], [ 0., 0., 0., 0., 0., 0., 108.]], dtype=float32)) ordembedding1: [[83, 104, 114, 105], [110, 105], [118, 97, 97], [115, 97, 110]] ordembedding2: [[83, 114, 105], [110, 105], [118, 97, 115, 97, 110]] Earth Mover Distance between two String syllable tensors: (0.3320339024066925, None, array([[ 83., 0., 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 104., 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 9., 105., 0., 0., 0., 0., 0., 0., 0.], [ 0., 1., 0., 0., 0., 0., 0., 0., 0., 104.], [ 0., 0., 0., 110., 0., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 105., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 118., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 97., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 97., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 11., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 7., 97., 6.]], dtype=float32))
Srinivasan Kannan
More... almost 6 years ago
878. (FEATURE) EventNet Python Parser - Events as Tensorflow Tensors - 20 November 2020 - related to 701 and all sections on EventNet Logical Clock, Video EventNet Tensor Products algorithm for Merit of Large Scale Visuals and Category Theory
Srinivasan Kannan
More... almost 6 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 878. (FEATURE) EventNet Python Parser - Events as Tensorflow Tensors - 20 November 2020 - related to 701 and all sections on EventNet Logical Clock, Video EventNet Tensor Products algorithm for Merit of Large Scale Visuals and Category Theory -------------------------------------------------------------------------------------------------------------------------------------------- This commit refactors EventNet/EventNet.py to define EventNet_Tensor_Parser() function which parses the EventNet Vertices and Edges text files and instantiates TensorFlow Tensor objects for each intraevent conversations (edges amongst actors in an event) and a holistic EventNet TensorFlow Tensor for interevent causality. EventNet abstraction helps analyzing Videos - each Frame is an Event, actor objects within a Frame event category interact by morphisms (intraevent) and Causality between Frames (interevent) is defined by edges or Functors between two Frame Event categories. More than anything else, EventNet conceptualizes a multidimensional logical time Tensor and postulates that time tensor could be decomposed to components. EventNet example has been redrawn by GraphViz and testlogs/EventNet.Tensor.log.20November2020 capture the EventNet Tensor instantiated as TensorFlow tf.Tensor() objects.
Srinivasan Kannan
More... almost 6 years ago
877. (FEATURE) Merit of Large Scale Visuals (LSVR) - TensorFlow Keras Backend - 18 November 2020
Srinivasan Kannan
More... almost 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 877. (FEATURE) Merit of Large Scale Visuals (LSVR) - TensorFlow Keras Backend - 18 November 2020 --------------------------------------------------------------------------------------------------------------------------------------------- This commit changes KERAS_BACKEND to TensorFlow for Video predictions by Keras and ImageNet. By this both Theano and TensorFlow are supported by NeuronRain Large Scale Visual Analytics. An example GoogleScholar and Google Search for query "NeuronRain Srinivasan Kannan" is analyzed by Keras-TensorFlow combine. TensorFlow has inbuilt support for tensors and Logical Clock EventNet Tensors, Video EventNet Tensors, People Profile Tensors and any multidimensional data could be translated to TensorFlow tf.Tensor() object. TensorFlow supports Keras models.
Srinivasan Kannan
More... almost 6 years ago
Updated Requirements.txt
Srinivasan Kannan
More... almost 6 years ago
876. (THEORY and FEATURE) People Analytics - People profiles as Tensors - Chaotic Hidden Markov Model of Tenure Transitions - TensorFlow integration for conversion of People profile data to Tensors - 17 November 2020 - related to 365,400,572,875 and all sections on Intrinsic merit of People, BKS Conjecture, Condorcet Jury Theorem, Margulis-Ruzzo Threshold, Berry-Esseen Central Limit Theorem and Stability of Interview formalized by LTF/TQBF/PTF
Srinivasan Kannan
More... almost 6 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 876. (THEORY and FEATURE) People Analytics - People profiles as Tensors - Chaotic Hidden Markov Model of Tenure Transitions - TensorFlow integration for conversion of People profile data to Tensors - 17 November 2020 - related to 365,400,572,875 and all sections on Intrinsic merit of People, BKS Conjecture, Condorcet Jury Theorem, Margulis-Ruzzo Threshold, Berry-Esseen Central Limit Theorem and Stability of Interview formalized by LTF/TQBF/PTF --------------------------------------------------------------------------------------------------------------------------------------------- 1.This commit invokes TensorFlow convert_to_tensor() function for converting any People profile data encoded as any Python object including JSON to a Tensor. 2.python-src/SocialNetworkAnalysis_PeopleAnalytics_ChaoticHMM.py which models tenure transitions of People profiles by Chaotic Hidden Markov Model Viterbi Trellis,now imports TensorFlow and TensorFlow I/O. 3.SocialNetworkAnalysis_PeopleAnalytics_ChaoticHMM.json Tenure Transition is read and the fields are converted to Tensors by convert_to_json() TensorFlow API e.g tenure organizations - academic and work - are encoded as TF tensor below: piplprofiletensor - states: tf.Tensor( [b'Schooling' b'DBHPS' b'PSGTech' b'BaaN-SSAGlobal' b'SunMicrosystems-Oracle' b'Krishna-iResearch' b'Verizon' b'webMethods-SoftwareAG' b'CMI-IIT-IMSc' b'GlobalAnalytics' b'Clockwork-PiQube' b'CloudEnablers'], shape=(12,), dtype=string) 4.TensorFlow function decode_as_json() might also be invoked but there are occassional JSON parsing errors. 5.Objective of NeuronRain People Analytics is to automate, do away with or minimize human errors in talent recruitment through machine learning models. Stability and Noise Sensitivity criterion of an interview TQBF/LTF/PTF derived earlier theoretically stipulates when a talent recruitment system by Question-Answering is better than majority voting and thus has enormous HR analytics ramifications. Goodness of majority voting - when group decision converges to good outcome for p-biased voters - is limited by Margulis-Ruzzo Threshold/Condorcet-Black-Ladha Jury Theorems for p-biased boolean functions. Combining the two, Stability criterion for interview TQBF (which is in IP=PSPACE) outclassing majority decision and Margulis-Ruzzo/CJT threshold for majority decision converging to 100% are thus inversely related - if Questions-Answers are improperly weighted (a <= 1.2056205488/n) for every possible LTF, Interview process is subverted and becomes unstable making BKS Conjecture False.
Srinivasan Kannan
More... almost 6 years ago
875. (THEORY and FEATURE) People Analytics - Alphabet Vectorspace Embedding Distance implementation - related to 752,874 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics - 15 November 2020
Srinivasan Kannan
More... almost 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------- 875. (THEORY and FEATURE) People Analytics - Alphabet Vectorspace Embedding Distance implementation - related to 752,874 and all sections on Embedding in vectorspace, Compressed Sensing, Social Network Analytics, Meaning Representation, Intrinsic Merit of People Profiles/HR Analytics - 15 November 2020 ------------------------------------------------------------------------------------------------------------------------------------------- 1. CompressedSensing.py has been changed to implement a new function alphabet_vectorspace_embedding_distance(self, embedding1, embedding2) which finds the euclidean distance between two string vector points (1-dimensional Tensor) embedded in alphabet space. 2. It uses the ord() builtin to compute ordinal values of the ascii or unicode strings. 3. alphabet_vectorspace_embedding_distance() is invoked from python-src/SocialNetworkAnalysis_PeopleAnalytics.py to compare two different name spellings in English of a Sanskrit name. 4. Strings of different lengths are left justified and padded by "#" fill before distance computation. 5. Distance computed is normalized - divided by length of the strings compared. 6. Logs python-src/testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.15November2020 show distance similarities amongst different name spellings in English of Sanskrit name which can be contrasted against Phonetic match rating and NYSIS Flight manifest name similarity algorithms as below: embedding1: ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#', '#', '#'] embedding2: ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#', '#', '#'] and ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] : 127.251719045363 Normalized Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#', '#', '#'] and ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] : 10.604309920446918 leftjustified: ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#'] embedding1: ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#'] embedding2: ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n'] Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#'] and ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n'] : 84.34453153583817 Normalized Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 's', '#'] and ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n'] : 8.434453153583817 leftjustified: ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n', '#', '#'] embedding1: ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] embedding2: ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n', '#', '#'] Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] and ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n', '#', '#'] : 103.24243313676794 Normalized Alphabet Vectorspace Embedding Distance between ['S', 'h', 'r', 'i', 'n', 'i', 'v', 'a', 'a', 's', 'a', 'n'] and ['S', 'r', 'i', 'n', 'i', 'v', 'a', 's', 'a', 'n', '#', '#'] : 8.603536094730662 7.Invocation of Recursive Lambda Function Growth Meaning Representation of a Professional Profile and Email context name parsing have been uncommented. Because of unfiltered proper nouns of organizations, few lambda functions are misleading. 8.Two phonetically different name spellings "Rasa" and "Raja" of same name (in Tamizh and Sanskrit respectively) are compared for euclidean distance similarity. 9.Syllable Hyphenation of name strings adds one more dimension and would create a 2-dimensional tensor embedding and distance between two 2D tensors could be more accurate e.g: "Srinivasan" - [[S,r,i,n,i],[v,a,s,a,n]] "Shrinivas" - [[S,h,r,i,n,i],[v,a,s]] "Shrinivaasan" - [[S,h,r,i,n,i],[v,a,a,s,a,n]] 10.For arbitrary dimensional people data as JSON tensors which include name,work experience,connections,academics etc., distance computation between Person1 and Person2 requires expanding dimensions of smaller tensor and computing the norm of Person1 - Person2 (Stackoverflow question - https://stackoverflow.com/questions/55196194/how-to-calculate-the-euclidean-distance-between-a-2-d-tensor-and-a-3-d-tensor)
Srinivasan Kannan
More... almost 6 years ago
874. (THEORY and FEATURE) People Analytics - People Profiles as Tensors - related to 752 and all sections on Embedding in vectorspace, Social Network Analytics, Psephology/Election analytics, Majority Voting, Intrinsic merit of People profiles - 13 November 2020
Srinivasan Kannan
More... almost 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 874. (THEORY and FEATURE) People Analytics - People Profiles as Tensors - related to 752 and all sections on Embedding in vectorspace, Social Network Analytics, Psephology/Election analytics, Majority Voting, Intrinsic merit of People profiles - 13 November 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1.python-src/SocialNetworkAnalysis_PeopleAnalytics.py has been changed to optionally choose between REST API calls to PIPL.com and reading a json file of people profile from any datasource including PIPL.com by boolean parameter loadfromjson in function pipldotcom_analytics() which reads People data in JSON from testlogs/SocialNetworkAnalysis_PeopleAnalytics.json 2.JSON objects are multidimensional which could be construed as points on a vectorspace. TensorFlow supports reading JSON format of Tensors as Protobufs (Read JSON string and return a Protobuf Tensor - https://www.tensorflow.org/api_docs/python/tf/io/decode_json_example). 3.Mapping people profiles onto Tensor points on vectorspace opens up possibility of whole gamut of Computational geometric People analytics - for example closest pair and clusters of people could be found and profiles could be iteratively ranked to get similarities. 4.Clustering of people profiles is crucial in analyzing voting patterns. 5.On a related note, NeuronRain implements a Set partition Electronic Voting Machine and every candidate bucket is a cluster of voters - Distance similarity (or Swing between consecutive polls) between any two Set Partition EVMs is measurable by Earth mover distance (Wasserstein distance).
Srinivasan Kannan
More... almost 6 years ago
Updated Requirements.txt
Srinivasan Kannan
More... about 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 873. (THEORY and FEATURE) Fibonaccian Search - Python 3.9.0 upgrade for Factorization - related to 872 - 5 November 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1. Fibonaccian search python implementation has been upgraded to Python 3.9.0 by autopep8 and 2to3-3.9 tools for potential invocation from Computational geometric factorization which is already Python 3 compatible. 2. Python 2.7 version of Fibonaccian search has been separately committed. 3. After this upgrade NeuronRain supports all three versions of Python 3 - Python 3.7.5, Python 3.8.5 and Python 3.9.0
Srinivasan Kannan
More... about 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 873. (THEORY and FEATURE) Fibonaccian Search - Python 3.9.0 upgrade for Factorization - related to 872 - 5 November 2020 ------------------------------------------------------------------------------------------------------------------------------------------ 1. Fibonaccian search python implementation has been upgraded to Python 3.9.0 by autopep8 and 2to3-3.9 tools for potential invocation from Computational geometric factorization which is already Python 3 compatible. 2. Python 2.7 version of Fibonaccian search has been separately committed. 3. After this upgrade NeuronRain supports all three versions of Python 3 - Python 3.7.5, Python 3.8.5 and Python 3.9.0
Srinivasan Kannan
More... about 6 years ago
Section 636 of NeuronRain Theory Drafts updated - Sorting of points on an algebraic curve - Convex segment decomposition
Srinivasan Kannan
More... about 6 years ago
872. (THEORY and FEATURE) Fibonaccian Search - alternative to binary search - related to all sections on Computational Geometric Planar Point Location Factorization, Geometric Search of BigData, Arithmetic Progressions and Ramsey coloring - 31 October 2020
Srinivasan Kannan
More... about 6 years ago
------------------------------------------------------------------------------------------------------------------------------------------- 872. (THEORY and FEATURE) Fibonaccian Search - alternative to binary search - related to all sections on Computational Geometric Planar Point Location Factorization, Geometric Search of BigData, Arithmetic Progressions and Ramsey coloring - 31 October 2020 ------------------------------------------------------------------------------------------------------------------------------------------- 1. Binary Search could be costly in some architectures because of division by 2. Fibonaccian search based on Fibonacci series avoids division and depends only on addition and subtraction. Note: Fibonaccian search is different from Fibonacci search to locate maximum of a unimodal function. 2. Binary search is central to Computational Geometric Factorization to locate a factor point within a segment/interval/pixel-array-polygon which is a sorted arithmetic progression of ordinate products. 3. Fibonaccian search optimization to replace Binary search in planar point location could significantly reduce constant involved in O(logN) search. 4. Both Binary search and Fibonaccian search are O(logN) while the search tree of the latter makes the difference - Left subtree of Fibonacci search is 1.618 (Golden ratio) times more skewed than Right subtree. 5. In terms of runtime efficiency Fibonaccian search lies between Binary Search and Interpolation search (which is O(loglogN) and assumes elements are within a known range)
Srinivasan Kannan
More... about 6 years ago
636. (THEORY) Computational Geometric Factorization - Parallel Planar Point Location of factors in Monotone Subdivision formed by Hyperbolic Arc Pixel Polygons with no necessity for Ray Shooting Queries, Van Der Waerden Ramsey Coloring of Sequences,Computational Algebraic Geometry - related to 506,601,605 and other sections on factorization - 12 February 2019, 18 February 2019, 20 February 2019, 22 September 2019, 20 October 2020
Srinivasan Kannan
More... about 6 years ago
Section 738 updated for Approximate factors and Local search optimization for exact Factorization
Srinivasan Kannan
More... about 6 years ago
869. (THEORY and FEATURE) Drone Obstacle Avoidance by 3D Navigation on Airspace defined by Road Geometry - related to 582, 867 and all sections on Motion Planning, GIS Analytics, Logical clocks, Large Scale Visuals Recognition, Drone Swarms - 2 October 2020, 6 October 2020
Srinivasan Kannan
More... about 6 years ago