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
------------------------------------------------------------------------------------------------------- 635. (THEORY and FEATURE) Intrinsic Merit of Texts, Ramsey Theory, Complement Diophantines - Ramsey number of text graph - related to 2,617,630,632 and all sections on Fame and Merit - 11 February 2019 ------------------------------------------------------------------------------------------------------- 1.It has been mentioned earlier that texts which are concatenations of words can be represented by *) Matrix Diophantine Equations *) and (255 or 511) coloring of texts by ASCII/Unicode alphabets (excluding space) which give rise to ramsey ordering by Van Der Waerden Theorem assuming texts of size N are integer sequences of locations 1-to-N colored by alphabets containing monochromatic arithmetic progressions of alphabet(color) locations. 2.Best known upper bound for Van Der Waerden Number for a least length sequence of N integers = W(r,k) for r-coloring of an integer sequence and arithmetic progression of least length k <= 2^2^(r^2^2^(k+9)) by [Timothy Gowers] 3.Previous upperbound for texts colored by alphabets of size 255 implies size N of a natural language text to have length k arithmetic progressions on alphabet locations <= 2^2^(255^2^2^(k+9)) which is huge even for small k. 4.Ramsey coloring R(r,s) of graphs is equivalent of Van Der Waerden sequence coloring which defines emergence of monochromatic cliques r or independent sets s (also known as Friends and Strangers theorem). 5.By definition of complementation in generic sense, k-coloring partitions a set into k disjoint subsets which can be represented by some diophantine (e.g Beatty functions) - Complementation is a size 2 partition. 5.In this commit, Ramsey R2 function is invoked from NetworkX API on textgraphs of natural language texts and maxclique-maxindependentset pairs are computed. 7.This is a less intensive intrinsic merit measure for a text and Van Der Waerden number computation has no mention in python libraries presently and is daunting too - SAT solvers are used to compute maxcliques in some algorithms. 8.Following this reasoning, strings of octave alphabets in music can be construed as 12-coloring (and more including majors and minors) of music as sequence and indicates inevitable emergence of order from chaos.
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Software Analytics and Scheduler Analytics - Backpropagation correction
Srinivasan Kannan
More... over 7 years ago
--------------------------------------------------------------------------------------------------------- 634. (FEATURE) Software Analytics and Scheduler Analytics - Backpropagation correction - 9 February 2019 --------------------------------------------------------------------------------------------------------- 1.DeepLearning_SchedulerAnalytics.py and DeepLearning_SoftwareAnalytics.py have been updated for correction to a serious bug in BackPropagation computation. 2.Earlier, BackPropagation Neural Network (BPNN) object for Software and Scheduler Analytics was getting created for every input training dataset element erroneously. 3.This has been remedied by instantiating BackPropagation class only once outside the loop and updating only the input layer and/or expected output layer for each element in input dataset iteration. 4.Iterations have been curtailed for concise logs. 5.By this change, weights are updated for every psutil per process proc object (Scheduler Analytics) or systemwide process statistics info (Software Analytics) thus learning weights of the multilayer perceptron from system performance. 6.logs for this have been committed to testlogs/DeepLearning_SchedulerAnalytics.log.9February2019 and testlogs/DeepLearning_SoftwareAnalytics.log.9January2019 7.After sufficiently large number of input training data, hopefully this BPNN would be able to predict outputs from input per-process and systemwide performance statistics data e.g 99% load average.
Srinivasan Kannan
More... over 7 years ago
(THEORY and FEATURE) GIS and Urban Planning Analytics, Convex Hull
Srinivasan Kannan
More... almost 8 years ago
--------------------------------------------------------------------------------------------------------- 633. (THEORY and FEATURE) GIS and Urban Planning Analytics, Convex Hull - 31 January 2019 --------------------------------------------------------------------------------------------------------- 1. ImageGraph_KerasTheano.py has been changed to include a new function for finding convex hull of set of points/pixels in an image 2. QHull ConvexHull() function from SciPy Spatial library - https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.ConvexHull.html - has been invoked in the new convex_hull() function defined for this purpose. 3. An example demographic GIS satellite image of Chennai Megapolis (Population density of Urban Agglomeration) from SEDAC website - http://sedac.ciesin.columbia.edu/mapping/popest/gpw-v4/ - has been analyzed to draw a convex hull around thickly populated regions and convex hull vertices are printed in terms of some area units (not sqkm or sqmiles). 4. OpenCV2 provides api for convex hull too - https://docs.opencv.org/2.4/doc/tutorials/imgproc/shapedescriptors/hull/hull.html 5. Convex Hull analytics of GIS imagery is helpful for Urban planners and sustainable development.
Srinivasan Kannan
More... almost 8 years ago
Theory - Intrinsic Merit of Social Profile Vertices, Experiential Merit
Srinivasan Kannan
More... almost 8 years ago
References - Factorization
Srinivasan Kannan
More... almost 8 years ago
(FEATURE and THEORY) Intrinsic Merit of Music, Exact Learning of DFAs, Minimum Description Length Merit
Srinivasan Kannan
More... almost 8 years ago
------------------------------------------------------------------------------------------------------- 632. (FEATURE and THEORY) Intrinsic Merit of Music, Exact Learning of DFAs, Minimum Description Length Merit - 29 January 2019 - related to all sections on Intrinsic Merit and Fame - related to 67,68,69,593, 630 ------------------------------------------------------------------------------------------------------- 1. New function audio_merit() has been defined in AudioToBitMatrix.py which invokes minimum_descriptive_complexity() to compute MDL for string of octave notes for a musical clip. 2.Source files MinimumDescLength.py and Entropy.py have been changed to include __main__ and rewritten to define member function minimum_descriptive_complexity() and some other error checks for floating point overflow. 3.An example musical MP4 clip excerpt of JS Bach has been analyzed. This masterpiece is featured in Douglas Hofstadter's Godel-Escher-Bach:An Eternal Golden Braid for its self-similar strange loops of notes (a mandelbrot fractal set) 4.This implementation thus is a culmination of 4 classes of intrinsic merits on Internet and WWW - Text, Video, Audio and People - defined in this draft. 5. Following are intrinsic merit implementations in this draft which expand on publications http://arxiv.org/abs/1006.4458, http://www.nist.gov/tac/publications/2010/participant.papers/CMI_IIT.proceedings.pdf and http://arxiv.org/abs/1106.4102 (Complement Diophantines are intrinsic merit measures too because text obtained by word concatenations can be represented by word diophantine equations and can be Ramsey 255-colored by alphabet locations thereby creating arithmetic progressions): 5.1.Textual merit is analyzed by Recursive Gloss Overlap and Recursive Lambda Function Growth algorithms - Graph Tensor Neuron Network Merit. 5.2 People Merit is analyzed by a mixture model of Intrinsic(Academic/Work/Interview/Contest/Examinations) and Experiential Mistake Bound Merit by a recursive mistake correction tree differential equation. 5.3 Video Merit is analyzed by EventNet/ImageNet Tensor Products Algorithm. 5.4 Audio Merit is defined theoretically as complexity of exact learning of DFAs on strings of notes and their minimization in [Dana Angluin] L* Algorithm Model and implemented by a less intensive Minimum Description Length merit on string of notes in this commit. Implementation of L* could be exponential and there are PAC learning approximate DFA learning algorithms. 6.Previous algorithms for 4 classes of intrinsic merit are theoretical approximations based on algorithmic graph theory, causality and logical time in clouds, learning theory and connectomes with no invocation of statistical tool and by no means the only way to analyze merit but one of them.
Srinivasan Kannan
More... almost 8 years ago
Theory - PRAM-NC Integer Factorization - Parallel Planar Point Location - References
Srinivasan Kannan
More... almost 8 years ago
Theory - Hardness Amplification of Majority, Theoretical EVMs
Srinivasan Kannan
More... almost 8 years ago
Theory - LSH/Separate Chaining based Theoretical EVMs, Mining patterns in voting, Vulnerabilities analysis - Reference
Srinivasan Kannan
More... almost 8 years ago
Theory - PRAM Planar Point Location for Factorization and PRAM lowerbounds - reference
Srinivasan Kannan
More... almost 8 years ago
(THEORY) Computational Geometric Factorization, Alternative Ray Shooting Algorithm, BQP and QNC, Pythagorean Triples, Possibility of decoherence from Quantum NC to Classical PRAMs, No cloning theorem
Srinivasan Kannan
More... almost 8 years ago
Theory - Intrinsic Merit of Music
Srinivasan Kannan
More... almost 8 years ago
(THEORY and FEATURE) Music Pattern Mining - Audio to Notes - Intrinsic Merit of Music
Srinivasan Kannan
More... almost 8 years ago
--------------------------------------------------------------------------------------------------------- 630. (THEORY and FEATURE) Music Pattern Mining - Audio to Notes - Intrinsic Merit of Music - 12 January 2019 - related to all sections on Intrinsic Merit and Fame - related to 67,68,69,593 --------------------------------------------------------------------------------------------------------- 1. Implementations in NeuronRain on Intrinsic Merit Versus Fame previously have so far concentrated only on Merit of Text (TextGraphs), Large Scale Visuals (Videos and Images - LSVR) and People (Interview/Examination/Contest based merit, Analytics of Social Network Profile Vertices). Merit of Audio or Complexity of Music has been defined theoretically earlier as a minimum size DFA or Turing Machine accepting musical notes as language which is a minimum description length measure similar to Kolmogorov complexity. 2.To this effect music_pattern_mining/AudioToBitMatrix.py has been changed to include a new function which converts an audio waveform to frequency domain by FFT. These frequencies are in turn mapped to a string of musical notes over octave alphabet - C,D,E,F,G,A,B,C (12 including fine divisions of notes denoted by #) - librosa API are invoked. 3.Essentially by mapping music to huge strings over octave alphabets the problem of music mining is made easier and all algorithms implemented for string mining in NeuronRain apply to music as well. 4.Since music is often related to creative genius, complexity of string of notes is an approximation of merit. 5.String of musical notes differ from usual text strings because of an inherent periodicity in music (self-similar fractal-like ascending and descending sequence of notes) 6.logs for an example audio to notes string has been committed to python-src/music_pattern_mining/testlogs/AudioToBitMatrix.log.AudioToNotes.12January2019
Srinivasan Kannan
More... almost 8 years ago
References - Program Analysis - Call Graphs formalized by Call strings and transfer functions for entry and exit
Srinivasan Kannan
More... almost 8 years ago
References - Program Analysis - Call Graph Analysis
Srinivasan Kannan
More... almost 8 years ago
References - Program Analysis - Call Graph Analysis in NeuronRain - Comparing Call Graphs
Srinivasan Kannan
More... almost 8 years ago
Complement Diophantine Equations over Complex Numbers and Algebraic/Arithmetic Circuits for them, Blum-Shub-Smale Turing Machines, Thue and Siegel Equations, Elliptic Curves, Mordell Conjecture and Theorem, ABC Conjecture, Riemann Zeta Function
Srinivasan Kannan
More... almost 8 years ago
Feature - Software Analytics - Hardcoded inputlayer replaced by Psutils - 8 January 2019 - Usecase
Srinivasan Kannan
More... almost 8 years ago
--------------------------------------------------------------------------------------------------------------- 628. (FEATURE) Software Analytics - Hardcoded inputlayer replaced by Psutils - 8 January 2019 - Usecase in 399 implemented --------------------------------------------------------------------------------------------------------------- 1.DeepLearning_SoftwareAnalytics.py in software_analytics/ has been updated and hardcoded process statistics for input layer of BackPropagation have been replaced by Psutil systemwide load percentages 2.This implements the usecase for software analytics mentioned earlier and finds the weights for input layer, hidden layer and output layer of BackPropagation multilayered perceptron. 3.Psutil provides cpu_percentage(), memory_percentage() and dist_usage() functions fields from which have been used as input layer of the perceptron. 4.Sampling loop has been limited to 2 but it can be any number. For each value of instantaneous system load (cpu, memory, disk IO) backpropagation iterates for sometime (limited to 10000) and finds the weights. 5.For any two successive load values of input layer at two time points t and t+delta, the weights of the perceptron fluctuate and should theoretically stabilize for significant number of load samples (which is presently 2) 6.Output layer has been set to a high value of 99% implying heavily thrashed system and input percentage values are divided by 10000 instead of 100 for probabilities to prevent overflow errors (NaN and Inf). 7.Logs have been committed to testlogs/DeepLearning_SoftwareAnalytics.log.8January2019
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines defined on vector spaces and subspaces and homorphisms between them
Srinivasan Kannan
More... almost 8 years ago
Theory - Grounded Cognition of Language - References
Srinivasan Kannan
More... almost 8 years ago
Theory - ThoughtNet Hypergraphs - References
Srinivasan Kannan
More... almost 8 years ago
References - Definability
Srinivasan Kannan
More... almost 8 years ago
Theory - People Analytics - Crowdsourcing
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines and Chinese Room Argument against strong AI
Srinivasan Kannan
More... almost 8 years ago
Feature - Vowelless String (De)Compression - Revamp
Srinivasan Kannan
More... almost 8 years ago