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
Theory - Complementary Beatty Functions over reals - Reference
Srinivasan Kannan
More... over 7 years ago
(FEATURE) People Analytics - Rank Correlation based Attrition Model of Tenure Histogram
Srinivasan Kannan
More... over 7 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 674. (FEATURE) People Analytics - Rank Correlation based Attrition Model of Tenure Histogram - 1 July 2019 --------------------------------------------------------------------------------------------------------------------------------------------- 1. People Analytics implementation has been updated to include a function to compute Correlation Coefficients of Statistical Dependence between vectors of Tenure Histograms. 2. Tenure Histogram is deemed as a set of vectors of the form (organization/designation, income, duration) which captures the crucial factors of a tenure 3. New function tenure_partition_rank_correlation() has been implemented to compute rank correlations between each pair of variables of a tenure: 3.1 Designations versus Remunerations 3.2 Designations versus Durations 3.3 Remunerations versus Durations 4.An example tenure histogram vectors of previous format has been analyzed (presently not parsed from profiles) for Kendall-Tau Rank correlation coefficient and logs are in testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.1July2019 5.Previous correlations can be interpreted as below: 5.1 Designations versus Remunerations - If positive or high, professional profile has increasing career graph along with increasing remunerations (expected). If negative or low, professional profile has increasing career graph vis-a-vis reduced remunerations (not expected - implies profile preferred more challenging roles than remuneration) and vice-versa (not expected - implies profile preferred better remuneration than designations) 5.2 Designations versus Durations - If positive or high, profile has increasing career graph vis-a-vis durations of tenure (expected). If negative or low profile has increasing career graph vis-a-vis decreasing durations per tenure (not quite expected - increased designations were not appealing to profile and person decided to leave) and vice-versa (not quite expected - decreasing responsibilities of the role were preferred by the profile which decided to stay longer) 5.3 Remunerations versus Durations - If positive or high, profile has increasing emoluments versus increasing durations per tenure (expected). If negative or low profile has increasing remunerations against decreasing durations per tenure (not quite expected - profile was more inclined to salary) and vice-versa (not quite expected - profile stayed longer per tenure despite decreasing benefits). 6. Previous interpretations facilitate inferences (in parentheses) of why attritions could have happened. 7. From logs following positive and negative correlations are found and can be interpreted based on (5): Kendall Tau Rank Correlations - Designations and Remunerations: tau= 0.7142857142857143 , pvalue= 0.0301587301587 Kendall Tau Rank Correlations - Designations and Durations: tau= -0.09759000729485331 , pvalue= 0.7612636496364197 Kendall Tau Rank Correlations - Durations and Remunerations: tau= -0.19518001458970663 , pvalue= 0.5434238636256696
Srinivasan Kannan
More... over 7 years ago
(THEORY) Computational Geometric Factorization, Ramsey Coloring and Arithmetic Progressions in the Tile Segments, Van Der Waerden Numbers - Update
Srinivasan Kannan
More... over 7 years ago
(FEATURE) People Analytics - Partition Rank Attrition Metric of Tenure Histogram
Srinivasan Kannan
More... over 7 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 673. (FEATURE) People Analytics - Partition Rank Attrition Metric of Tenure Histogram - 29 June 2019 -------------------------------------------------------------------------------------------------------------------------------------------- 1.People Analytics implementation has been updated to compute partition rank described previously (Crank of a Partition - Freeman Dyson - https://en.wikipedia.org/wiki/Crank_of_a_partition) of tenure histogram set partition of a person's professional and academic profile. 2.Recursive Lambda Function Growth algorithm has been applied to text of the profile again.Closeness and Betweenness Centralities rankings are quite relevant than other Centrality measures. 3.Logs for this have been committed to testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.29June2019 x
Srinivasan Kannan
More... over 7 years ago
(FEATURE and THEORY) Intrinsic Merit of People/Professional Profiles - People Analytics - Tenure Histogram
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 672. (FEATURE and THEORY) Intrinsic Merit of People/Professional Profiles - People Analytics - Tenure Histogram - 28 June 2019 ------------------------------------------------------------------------------------------------------------------------------------------ 1.People Analytics implementation has been updated to print Work and Academic Tenure stints as array (histogram or integer partition) 2.This essentially makes tenures as a set partition analytics problem because job and academic stint time durations can be picturized as histograms. 3.Tenures represented as Set Partitions can be quantified by set partition measure of partition rank: Length of largest part in partition - Number of parts in partition 4.In the context of People Analytics, previous rank measure is quite applicable practically as attrition metric because: 4.1 High partition rank implies the professional profile has spent large duration in an organization or number of tenure switches are less (number of parts) 4.2 Low Partition rank implies the professional profile has low duration per organization or high number of tenure switches 5.Tenure Partition can be printed as Ferrer or Young Diagram and size of Durfee Square of this tenure partition is also a measure of average time spent per organization and thus an attrition metric 6.Recursive Lambda Function Growth wordnet error has been resolved (pywsd is not used) and import uncommented. 7.logs for this have been committed to testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.28June2019
Srinivasan Kannan
More... over 7 years ago
(THEORY and FEATURE) Intrinsic Merit of Music - MFCC analysis and zero crossing rate
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 671. (THEORY and FEATURE) Intrinsic Merit of Music - MFCC analysis and zero crossing rate - 27 June 2019 ------------------------------------------------------------------------------------------------------------------------------------------ 1.AudioToBitMatrix.py MFCC function has been updated to scale MFCCs by sklearn and print mean and variance of MFCCs of an audio/music waveform 2.zero crossing rate of waveform is also printed which is measure of modulations in music. 3.One of the ways to measure Music merit and Music Genre or Emotion Classification is by clustering music waveforms by distance similarity of MFCCs. 4.MFCC similarity can be used as recommender system where a listener is recommended music of similar MFCCs based on historic listening habits. 5.MFCCs of music waveform drawn as images (e.g by librosa.display.specshow()) can be compared for similarities.
Srinivasan Kannan
More... over 7 years ago
(THEORY) Computational Geometric NC-PRAM Factorization, Tile Summation Ray Shooting Queries and Gamma approximation, Iterated addition for sum of tile segment lengths
Srinivasan Kannan
More... over 7 years ago
(THEORY) Computational Geometric Factorization, Ramsey Coloring and Arithmetic Progressions in the Tile Segments, Van Der Waerden Numbers
Srinivasan Kannan
More... over 7 years ago
(THEORY) Computational Geometric Factorization - Average case Sequential Optimization, Stirling Approximation of Gamma Function, Sum of binary search times of each tile and some observations - correction
Srinivasan Kannan
More... over 7 years ago
(THEORY) Computational Geometric Factorization - Average case Sequential Optimization, Stirling Approximation of Gamma Function, Sum of binary search times of each tile and some observations
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Social Network Analysis, People Analytics, PDF file parsing
Srinivasan Kannan
More... over 7 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 667. (FEATURE) Social Network Analysis, People Analytics, PDF file parsing - 13 June 2019 --------------------------------------------------------------------------------------------------------------------------------------------- 1. Changed SocialNetworkAnalysis_PeopleAnalytics.py to parse PDF files of social network profiles by PyPDF2 which was not working earlier. 2. PyPDF2 expects PDF files to be written by PDF writer or otherwise blank text is extracted 3. As example CV of self written by LaTeX2PDF has been parsed to print profile text to testlogs/SocialNetworkAnalysis_PeopleAnalytics.log.13June2019 4. Recursive Lambda Function Growth import has been commented because of WordNet error: Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/local/lib/python2.7/dist-packages/wn/__init__.py", line 11, in <module> from wn.info import InformationContentSimilarities File "/usr/local/lib/python2.7/dist-packages/wn/info.py", line 7, in <module> from wn.reader import parse_wordnet_ic_line File "/usr/local/lib/python2.7/dist-packages/wn/reader.py", line 22 offset, lexname_index, pos, n_lemmas, *the_rest = columns_str.split() ^ SyntaxError: invalid syntax 5.LinkedIn PDF Profile has been updated by linkedin pdf export
Srinivasan Kannan
More... over 7 years ago
(THEORY) Approximation of Majority Voting, Streaming majority, Forecasts, Bertrand Ballot Theorem, Theoretical EVMs, Set Partitions
Srinivasan Kannan
More... over 7 years ago
Theory - Segment Trees and Point location - References
Srinivasan Kannan
More... over 7 years ago
Theory - Planar Point Location Algorithms - References
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Named Entity Recognition - Conditional Random Fields - Skew Normal Distribution Update
Srinivasan Kannan
More... over 7 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 665. (FEATURE) Named Entity Recognition - Conditional Random Fields - Skew Normal Distribution Update - 29 May 2019 -------------------------------------------------------------------------------------------------------------------------------------------- 1.Miscellaneous changes in NamedEntityRecognition_HMMViterbi_CRF.py for corrections to skew normal distribution which shifts mean of the distribution for each part of speech to the right by some constant multiples. 2.Skew normal distribution mean shift for each state (part of speech) is shown in NamedEntityRecognition_HMMViterbi_CRF.29May2019.png 3.Motivation for skew normal distribution mean shift from left to right for emission probabilities is: For some part of speech state, successive word observations in the sentence receive more weightage corresponding to the peak value at mean i.e one word in the sentence peaks at mean and is more likely candidate observation emitted for respective Part of Speech state. 4.In this example 5 words in the sentence peak for 5 part of speech states. 5.logs for this commit have been committed to testlogs/NamedEntityRecognition_HMMViterbi_CRF.log.29May2019
Srinivasan Kannan
More... over 7 years ago
Updated NeuronRain usecases
Srinivasan Kannan
More... over 7 years ago
Theory - Theoretical EVMs - Voter Received Encrypted Paper Audit Trail (VREPAT)
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Spark 2.4 Structured Streaming - Windowed Stream
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------- 664. (FEATURE) Spark 2.4 Structured Streaming - Windowed Stream - 21 May 2019 ------------------------------------------------------------------------------------------------------- 1.SparkGenericStreaming.java in java-src/bigdata_analytics/spark_streaming has been changed to do windowed structured stream if URL socket is false i.e data is read from plain socket stream. 2.New boolean flag "windowed" has been added to the java class as static member based on which Windowed streaming spark code is executed. 3.Windowed streaming spark code fragment uses readStream()/writeStream() facility in Spark to read data written over socket. 4.Netcat utility creates the socket server in port 8080. 5.Java lambda functions in FlatMapFunction tokenize the incoming stream of lines to words and organize them as word-timestamp rows dataset. 6.Next groupBy() parses word-timestamp dataset from previous lambda invocation and creates structured stream of tables of columns slidingwindow_timestamp-word-count. 7.Logs in testlogs/SparkGenericStreaming.log.21May2019 show dataframe tables for both urlsocket=false(windowed=true) and urlsocket=true boolean option settings. 8.CLASSPATH for compilation of SparkGenericStreaming.java (javac SparkGenericStreaming.java) has been updated to: CLASSPATH=/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/spark-streaming_2.11-2.4.0.jar:.:/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/scala-library-2.11.12.jar:/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/spark-sql_2.11-2.4.0.jar:/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/spark-core_2.11-2.4.0.jar:/media/Ubuntu2/jdk1.8.0_171/lib/jsoup-1.11.3.jar:/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/spark-catalyst_2.11-2.4.0.jar:/media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/jars/scala-reflect-2.11.12.jar:/media/Ubuntu2/jdk1.8.0_171/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 9.sparkgenericstreaming.jar is packaged as bin/jar cvf sparkgenericstreaming.jar *.class 10.SparkGenericStreaming is executed as: /media/Ubuntu2/spark-2.4.0-bin-hadoop2.7/bin/spark-submit --jars /media/Ubuntu2/jdk1.8.0_171/lib/jsoup-1.11.3.jar --class SparkGenericStreaming --master local[2] sparkgenericstreaming.jar "https://twitter.com/search?f=tweets&vertical=news&q=Chennai&src=typd"
Srinivasan Kannan
More... over 7 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 7 years ago
Feature - Named Entity Recognition - Conditional Random Fields
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 663. (FEATURE) Named Entity Recognition - Conditional Random Fields - Update - 16 May 2019 ------------------------------------------------------------------------------------------------------------------------------------------ 1.NamedEntityRecognition_HMMViterbi_CRF.py has been changed to removed hardcoded PoS tagged sentences and instead to read states and observations from two files: NamedEntityRecognition_HMMViterbi_CRF.states and NamedEntityRecognition_HMMViterbi_CRF.observations which have the semi-PoS tagged sentence PoS states and sentence word observations respectively. 2.Emission Probabilities for each state are computed by a skew normal distribution which moves the mean from left to right of the sentence as states progress. 3.This skew normal distribution gives more weightage to respective state than others while each state is observed. 4.An example log in testlogs/NamedEntityRecognition_HMMViterbi_CRF.log.16May2019 shows the skewing of probability weights from left to right of the sentence 5.For example, Name-Verb-Object PoS tag is weighted as : First 1/3 of sentence has more weightage for emission probabilities of "Name",second 1/3 of sentence has more weightage for emission probabilities of "Verb" and third 1/3 of sentence has more weightage for emission probabilities of "Object"
Srinivasan Kannan
More... over 7 years ago
Theory - EventNet Tensor Products, Video Merit, Like Versus Merit
Srinivasan Kannan
More... over 7 years ago
Theory - Theoretical EVMs, Set Partitions, Audit Trails - References
Srinivasan Kannan
More... over 7 years ago