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
Feature - ConceptNet 5.6 Least Common Ancestor Path between Concepts rewrite
Srinivasan Kannan
More... over 8 years ago
---------------------------------------------------------------------------------------------------------- ConceptNet 5.6 Client Update - Common Ancestor Algorithm rewrite for path between concepts ---------------------------------------------------------------------------------------------------------- 1. ConceptNet 5.6 provides a REST endpoint /query for finding connections between two concepts which has been wrapped by a new function conceptnet_path() 2. Common Ancestor Algorithm for finding distance betwen two concepts has been rewritten to create a NetworkX subgraph from the edges connecting two concepts. Function conceptnet_distance() has been renamed to conceptnet_least_common_ancestor_distance() 3. Loop iterating over cartesian product has been split into 2 loops which grow from two opposite directions (concept1, --- ) and (--- ,concept2) and intersect at common ancestors. 4. Edges recursively seen by /related REST query are collected in a list and a NetworkX DiGraph is created from it which has all paths between concept1 and concept2 5. Shortest path on this subgraph yields the shortest distance and path between concept1 and concept2 and thus passes via the least common ancestor between two concepts 6. Logs for this have been captured in testlogs/ConceptNet5Client.log.19July2018 which show conceptnet_least_common_ancestor_distance() computes deeper paths than /query
Srinivasan Kannan
More... over 8 years ago
Theory and Feature - Computational Geometric Factorization, Solution to Pell Diophantine Equation, Ray Shooting Queries
Srinivasan Kannan
More... over 8 years ago
-------------------------------------------------------------------------------------------------------------- Computational Geometric Factorization and its Pell Diophantine (Complement Diophantine of Prime Valued Polynomials) - Updates -------------------------------------------------------------------------------------------------------------- 1. DiscreteHyperbolicFactorizationUpperbound_TileSearch_Optimized.py has been updated to persist the factors found to a JSON file (json.dump() and json.load()). 2. Aggregating factors found by multiple Spark executors requires defining a new Spark Broadcast Accumulator class which appends a factor to already existing accumulator value globally 3. complement.py has been updated for solving both directions of Pell-Factoring (for D=1 and N > 1): 3.1 Pell Equation to Factoring: Solves Pell equation x^2-y^2=N for D=1 and finds factors (x+y) and (x-y) in exponential time (or between polynomial and exponential - http://www.math.leidenuniv.nl/~psh/ANTproc/01lenstra.pdf) 3.2 Factoring to Pell Equation: Computes all factors (p,q) of N by PRAM Computational Geometric Factorization and finds x and y in Pell Equation as: x=(p+q)/2 and y=(p-q)/2. This solves a special case of Pell Equation in classical NC for D=1 (for generic, only quantum polynomial time algorithms are known so far) 4. Logs for these have been committed to testlogs/DiscreteHyperbolicFactorizationUpperbound_TileSearch_Optimized.JSONPersistence.18July2018 and testlogs/complement.PellEquation.SolvedByPRAMNCFactoring.18July2018 5. Ray Shooting Oracle shows some exact and close ray shooting of factors. For simple Hardy-Ramanujan ray shooting factorization requires only O(loglogN) + O(log(N/loglogN)*log(N/loglogN)) + O(log(N/loglogN)) time to find a factor because prime factors are O(loglogN) and distance between any two prime factors is O(N/loglogN). 6. Hyperbolic arc bow for this x-axis distance is sufficient to locate a factor. K-mergesort/Segment-tree of this subset is O(log(N/loglogN)*log(N/loglogN)) and binary search is O(log(N/loglogN)).
Srinivasan Kannan
More... over 8 years ago
Feature - Spark Structured Streaming - Timestamps in DataFrame - Spark 2.3 Upgrade
Srinivasan Kannan
More... over 8 years ago
---------------------------------------------------------------------------------------------------------------- 576. (FEATURE) Spark Streaming - Java 1.8 - Spark Java Upgrade to 2.3 and Some Structured Streaming Enhancements - 17 July 2018 ---------------------------------------------------------------------------------------------------------------- 1. Spark Generic Streaming Java Implementation has been compiled on Spark 2.3 upgraded from Spark 2.1 and sparkgenericstreaming.jar has been repackaged 2. JavaPairDStream code which was commented earlier has been uncommented to return Word Count instead of Words 3. Structured Streaming usecase example in https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html for website advertisement analytics has been implemented by computing timestamp of each word occurrence along with its count 4. Word.java has been changed for doing get/set of Time,Count and Word 5. DataFrame schema has been changed to Word, Count and Timestamp instead of just Word. 6. This Structured Streaming DataFrame prints periodically changing tables of word occurrences, frequencies and timestamp as a tabular column. 7. JavaPairDStream still does not have Java Pair support but only scala.Tuple2.
Srinivasan Kannan
More... over 8 years ago
Theory - Fame Versus Merit - Sybils - Margulis-Russo Threshold
Srinivasan Kannan
More... over 8 years ago
Theory - Computational Geometric Factorization - Wavelet Trees
Srinivasan Kannan
More... over 8 years ago
Theory - Computational Geometric Factorization by Parallel Construction of Wavelet Trees
Srinivasan Kannan
More... over 8 years ago
Feature - Deep Learning - Reinforcement Learning - Recommender Systems - Graph Tensor Neuron Network Instrinsic Merit
Srinivasan Kannan
More... over 8 years ago
Feature - Facebook Streaming Analytics - Comments
Srinivasan Kannan
More... over 8 years ago
-------------------------------------------------------------------------------------------------- Streaming Facebook Analytics - Comments -------------------------------------------------------------------------------------------------- For each wall post, comments object is retrieved by requests and printed
Srinivasan Kannan
More... over 8 years ago
Theory - Fame Versus Merit, Margulis-Russo Threshold and Least Energy definition of Intrinsic Merit
Srinivasan Kannan
More... over 8 years ago
Theory - Intrinsic Merit Function - Fame and Merit
Srinivasan Kannan
More... over 8 years ago
Feature - Facebook Streaming Analytics
Srinivasan Kannan
More... over 8 years ago
------------------------------------------------------------------------------------------------------------- Streaming Facebook Analytics - Exhaustively print all wall posts of a user ------------------------------------------------------------------------------------------------------------- 1. Loop has been added to Streaming_FacebookData.py for printing all pages of user wall posts in json 2. logs for this have been committed to testlogs/Streaming_FacebookData.log.6July2018
Srinivasan Kannan
More... over 8 years ago
Theory and Feature - Computational Geometric Factorization, Ray Shooting Query Oracles, ABC Conjecture
Srinivasan Kannan
More... over 8 years ago
Theory - Reference - Parallel Augmented Maps and Interval Trees
Srinivasan Kannan
More... over 8 years ago
Theory - Correction to SDP Least Squares Approximation SAT Solver
Srinivasan Kannan
More... over 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 8 years ago
Theory and Feature - Semidefinite Programming and Least Squares Approximate SAT Solver
Srinivasan Kannan
More... over 8 years ago
------------------------------------------------------------------------------------------------------- (THEORY and FEATURE) SAT Solver - Discrete Fourier Transform (FFT) of the real assignment, Parseval/Plancherel Theorems, Semidefinite Programming Approximation and Least Squares ------------------------------------------------------------------------------------------------------- 1. SAT Solver has been updated to compute Fast Fourier Transform of the real assignment points for each random 3SAT. 2. Discrete Fourier Transform of set of points is: X(N) = Sum(x(n) * e^(2*pi*n*k/N)), k=0,1,2,...,N-1 X(i) are frequencies and x(i) are discrete points in real assignment vector computed from Least Squares 3. Mapping the real assignment points to frequency domain enables finding periodicity within the real assignments and applying Parseval/Plancherel Theorems (https://en.wikipedia.org/wiki/Discrete_Fourier_transform) 4. Parseval Theorem equates the inner product of two vectors to inner product of Fourier Transform Frequencies. 5. Inner product of two assignment vectors (exact binary and real approximate - for values of variables) is a measure of distance between the exact (if known) and approximate assignments 6. Least Square Approximation solves the system of linear equations AX = B for random 3SAT while Semidefinite Programming (SDP) solves the Convex Optimization Problem: maximize/minimize: Double_Summation(a(i,k)*xi * yk) Subject to Constraints: Double_Summation(cl*xi*yk) < bl 7. In other words, SDP solves: maximize/minimize: A*X Subject to Constraints: C*X < B X is a Positive Semidefinite Gramian Matrix having entries from inner product space of vectors (or v*X*v^T is non-zero for real vectors v) 8. [Goemans-Williamson] MAXCUT Approximation algorithm - http://www-math.mit.edu/~goemans/PAPERS/maxcut-jacm.pdf -Journal of the Association for Computing Machinery,vol. 42, No.6, November1995 - solves MAXCUT in ~88% by SDP. 9. Any symmetric matrix can be approximated to nearest positive semidefinite matrix (e.g cov_nearest - http://www.statsmodels.org/dev/generated/statsmodels.stats.correlation_tools.cov_nearest.html) 10. There is an SDP variation of least squares, Semidefinite Least Squares, which solves the convex program: minimize: |X-C|^2 Subject to: AX=B for positive semidefinite matrix X. which is similar to least squares solution for variable vector X. X is positive semidefinite because for any real vector v and some assignment to X, v*X*vT is positive or zero. Minimizing |X-C|^2 for some binary assignment C finds closest approximation of X to binary step function characterized by C. (Example: https://ljk.imag.fr/membres/Jerome.Malick/Talks/08-louvain.pdf) 11.logs for 2000 variables and alpha=1 have been committed to testlogs/ which show usual MAXSAT approximation ratio convergence to >= 97%
Srinivasan Kannan
More... over 8 years ago
Feature - Streaming Facebook Analytics
Srinivasan Kannan
More... over 8 years ago
------------------------------------------------------------------------------------------------------------ Streaming Analytics - Facebook Graph ------------------------------------------------------------------------------------------------------------ 1.Streaming_FacebookData.py has been parametrized to accept user name as commandline argument $python3.4 Streaming_FacebookData.py <facebook_user> 2.Wall Posts and Friends of a user are obtained by facebook-sdk REST 3.Friends generator object is iterated and printed
Srinivasan Kannan
More... over 8 years ago
Feature-Facebook streaming analytics
Srinivasan Kannan
More... over 8 years ago
-------------------------------------------------------------------------------------------------------------- Streaming Analytics - Facebook Graph API - Python 3.4 and facebook-sdk -------------------------------------------------------------------------------------------------------------- 1.Similar to Streaming_<datasource>.py for LinkedIn and Twitter, Streaming_FacebookData.py has been added to NeuronRain AsFer repositories in SourceForge,GitHub and GitLab. 2.This imports facebook-sdk for python 3.4. This is the first python implementation on python 3.4. 3.This upgrade to python 3.4 from python 2.7 is necessitated because: 3.1 facebook-sdk on python 2.7 requires python 2.7.9 3.2 But importing facebook-sdk on Python 2.7.9 raises errors related to _ssl and _hashlib (md5,sha etc.,) which are not by default supported by python 2.7.9 3.3 Enabling SSL flags in Modules/Setup.dist for python 2.7.9 non-system builds also does not resolve this error and ssl is not built along with python 3.4 Facebook sdk supports only python 3.4 and 3.7 in 3.x 3.5 Python3.4 has been installed from a non standard PPA 3.6 VirtualEnv, EasyInstall3, PIP also do not solve the above SSL import issue 3.7 In python3.4 too, lot of facebook-sdk dependencies have to be git-cloned,built and installed (requests, idna, certifi, chardet etc.,) for facebook import to work 4. Example temporary access token has been created from from https://developers.facebook.com/tools/explorer/ 5. Simple name and id REST query has been done using requests and logs committed to testlogs/
Srinivasan Kannan
More... over 8 years ago
Theory - Parallel Construction of Segment Tree for Computational Geometric Factorization
Srinivasan Kannan
More... over 8 years ago
Theory - Brain Graphs or Connectome analysis - Reference
Srinivasan Kannan
More... over 8 years ago
Theory - Infeasibilit of Sequential Computational Geometric Factorization
Srinivasan Kannan
More... over 8 years ago
Theory - Exact Learning of a Universal CNF for intrinsic merit
Srinivasan Kannan
More... over 8 years ago