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
Updated BestPractices.txt
Srinivasan Kannan
More... about 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... about 8 years ago
Theory - Binary Least Squares - Reference
Srinivasan Kannan
More... about 8 years ago
Theory - Wavelet Tree for 2-dimensional range search for factors
Srinivasan Kannan
More... about 8 years ago
Feature - Music/Audio Pattern Mining - Jensen-Shannon Divergence of Probability Distributions of two audio signals
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------- Music/Audio Pattern Mining - Jensen-Shannon Divergence of Probability Distributions of two audio signals ------------------------------------------------------------------------------------------------------------- 1.audio_features() in AudioToBitMatrix.py returns the probability histogram and bins in a tuple 2.JensenShannonDivergence.py has been rewritten to compute normalization inline (normalize() function has been removed), import audio functions from AudioToBitMatrix.py and compute probability histograms for two different audio files. 3.logs for Jensen-Shannon Divergence distance between two audio files have been committed to testlogs/JensenShannonDivergence.log.8September2018 (For same audio, distance is zero)
Srinivasan Kannan
More... about 8 years ago
Feature - Histogram/Probability Distribution from Audio
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ Music/Audio Pattern Mining - Histogram and Probability Distribution Function extraction (PDF) ------------------------------------------------------------------------------------------------------------ 1.New function audio_features() has been added for analyzing features in audio/music signal 2.Considering audio as time series there are 2 standard ways to extract probability distribution from time series: 1) Histogram 2) Bandt and Pompe - [Andres M. Kowalski,Maria Teresa Martin,Angelo Plastino and George Judge] - www.mdpi.com/1099-4300/14/10/1829/pdf 3.audio_features() invokes histogram() function and extracts PDF from NumPy by setting density=True 4.Comparing two music signals amounts to finding Jensen-Shannon Divergence of two probability distributions 5.Histogram for an example audio signal is printed in testlogs/AudioToBitMatrix.log.7September2018
Srinivasan Kannan
More... about 8 years ago
Theory - Machine Translation, TextGraph Isomorphism
Srinivasan Kannan
More... about 8 years ago
Theory - Horn3SAT and Least Squares SAT Solver
Srinivasan Kannan
More... about 8 years ago
Theory - Wavelet Tree for Computational Geometric Factorization - Reference
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - Software Similarity Detection
Srinivasan Kannan
More... about 8 years ago
Updated BestPractices.txt
Srinivasan Kannan
More... about 8 years ago
Theory - Wavelet Trees, Pixelated Hyperbolic Curve Tile Summation and Factorization
Srinivasan Kannan
More... about 8 years ago
Theory - Wavelet Trees, Pixelated Hyperbolic Curve Tile Summation and Factorization
Srinivasan Kannan
More... about 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... about 8 years ago
Feature - Audio Pattern Mining - librosa time series
Srinivasan Kannan
More... about 8 years ago
--------------------------------------------------------------------------------------------------------------- Audio/Music Pattern Mining - Audio to Time Series --------------------------------------------------------------------------------------------------------------- 1.Similar to image_pattern_mining/ImageToBitMatrix.py, new file AudioToBitMatrix.py has been committed to repository for converting an audio file (e.g MP3, .wav) to time series array 2.AudioToBitMatrix.py imports librosa (audioread) Python library which returns waveform time series floating point array and sampling rate. 3.audio-to-bitmatrix function has optional arguments to specify if the time series is binary or float (defaults to float) and duration of the timeseries (defaults to full) 4.Binary time series invokes bin() on each time series point 5.An example audio MP3 file has been converted to time series and logs have been committed to testlogs/AudioToBitMatrix.log.24August2018
Srinivasan Kannan
More... about 8 years ago
Reference - Burrows-Wheeler Transform
Srinivasan Kannan
More... about 8 years ago
Theory - Burrows-Wheeler Transform and K-MergeSort for Factorization
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - L1 Norm Regularized Least Squares SAT Solver and gesv()
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------- SAT Solver - CVXOPT - L1-norm Regularized Least Squares Quadratic Program l1regls() and gesv() ------------------------------------------------------------------------------------------------------------- 1.This commit includes a special SAT Solver in CVXOPT documentation example l1regls() [https://cvxopt.org/examples/mlbook/l1regls.html] which solves the L1-norm regularized Least Squares Quadratic Program(QP): minimize error |AX-B|^2 + L1Norm(|X|) which is the most accurate Least Square Solution. Geometric intuition and preference of L1 norm for least squares is depicted in Q&A: https://www.quora.com/Why-is-L1-regularization-better-than-L2-regularization-provided-that-all-Norms-are-equivalent
Srinivasan Kannan
More... about 8 years ago
-------------------------------------------------------------------------------------------------------------- Recursive Lambda Function Growth - DOT file creation for Text Graph and density() correction -------------------------------------------------------------------------------------------------------------- 1.RecursiveLambdaFunctionGrowth.py has been changed to import write_dot() for pydot and dot file for Text Graph created has been written to RecursiveLambdaFunctionGrowth.dot for standardizing all graph implementations in NeuronRain to write out DOT file readable by Spark GraphX, NetworkX etc., 2.Bug in Graph Density computation has been fixed to invoke networkx.classes.function.density() which was earlier incorrectly printing the object ref 3.logs have been committed to testlogs/RecursiveLambdaFunctionGrowth.log.22August2018
Srinivasan Kannan
More... about 8 years ago
Feature - Program Analysis - Call Graph and CFG Graph Mining
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------- Program Analysis + Software Analytics - Graph Mining of Call graph and CFG DOT files ------------------------------------------------------------------------------------------------------------- 1.GraphMining_GSpan.py has been changed to check for an error condition of number of vertices per edge 2.New function graph_mining() has been included in software_analytics/CyclomaticComplexitySparkMapReducer.py for invocation of Graph Mining GSpan implementation to print frequent subgraph patterns in SATURN CFG and Valgrind/Callgrind/KCachegrind graph DOT files 3.logs for this have been committed to software_analytics/testlogs/CyclomaticComplexitySparkMapReducer.log.GraphMining.20August2018 4.This formally unifies the two ends: Program Analysis and Software Analytics (Deep Learning and Graph Mining) 5.Frequent Subgraphs in DOT files of Call graph and Control Flow Graph(CFG) extract static patterns in underlying software running on a system while the FlameGraph/Perf Graphics DeepLearning Analytics extract runtime patterns 6.Theoretically, if SAT for a Program is created by SATURN, Least Square SAT Solver in NeuronRain could be model checker for the Program
Srinivasan Kannan
More... about 8 years ago
Feature - Software Analytics - Program Analysis - Valgrind/Callgrind/KCachegrind Call graph analytics
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------- 590. (FEATURE) Program Analysis + Software Analytics - Valgrind/Callgrind/KCachegrind Call Graph DOT file creation and its Cyclomatic Complexity - 18 August 2018 ------------------------------------------------------------------------------------------------------------- 1. Valgrind/Callgrind are the standard tools for call graph profiling and memory leak check of userspace code. 2. Cyclomatic Complexity is already computed for DOT files created by SATURN Program Analyzer for Linux Kernel by VIRGO SATURN Program Analysis Driver. For userspace call graph DOT files are created by Kcachegrind complementing SATURN in userspace. 3. Valgrind for an example Spark program (Factorization) is invoked as: valgrind --tool=callgrind -v --dump-every-bb=10000000 /media/Ubuntu2/spark-2.3.1-bin-hadoop2.7/bin/spark-submit ../DiscreteHyperbolicFactorizationUpperbound_TileSearch_Optimized.py 819289 4. callgrind.out files are written to: callgrind.DiscreteHyperbolicFactorization_TileSearch_Optimized.out.14531 suffixed by process id. 5. callgrind.out file is viewed by kcachegrind (requires KDE) and call graph is exported as DOT and image files in: 5.1 kcachegrind_callgraph_DiscreteHyperbolicFactorization_TileSearch_Optimized.png 5.2 kcachegrind_callgraph_DiscreteHyperbolicFactorization_TileSearch_Optimized.dot 6. CyclomaticComplexitySparkMapReducer.py has been updated for doing Spark wordcount of both SATURN and Valgrind/Callgrind/KCachegrind DOT Control Flow Graph / call graph files and print Cyclomatic Complexity. 7. This leverages on already implemented Convolution Net Deep Learning of Graphic Performance Patterns by Perf and FlameGraph. 8. Essentially Program Behaviour Analysis in userspace+kernelspace reduces to Graph Mining of these DOT graphs.
Srinivasan Kannan
More... about 8 years ago
Theory - Cochrane-Dressler-ABC Conjecture ray shooting queries for factorization and S.S.Pillai Conjecture
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - SAT Solver CVXOPT gesv() invocation
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ SAT Solver - CVXOPT - gesv() least squares ------------------------------------------------------------------------------------------------------------ 1. CNFSATSolver.py has been changed to import gesv, getrs, sysv functions for least squares in cvxopt 2. Each of these were tried for 10 variables and 10 clauses combinations and gesv() has the highest accuracy observed so far (~99% to 100%) in any of the least squares algorithms. 3. logs for 500 random 3SAT instances have been committed to testlogs/CNFSATSolver.10variablesAlpha1.0.CVXOPT.log.16August2018 4. Caveat: Though the number of variables/clauses are less which could cause repetitive choice of same 3SAT similar randomness did not attain 99% accuracy for other least squares algorithms earlier for similar variable-clause combination. 5. gesv() also throws ArithmeticError for non full-rank matrices and logs for gesv() are filtered-out random 3SAT instances which are full rank. This leaves some random 3SAT instances unsolved. 6. try...except has been changed to print exception stacktrace for ArithmeticError
Srinivasan Kannan
More... about 8 years ago