Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 4 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
Uncommented NetworkX matplotlib graph plotting to get a graph for a sample document definition graph. Logs and PNG file for this classification have been added to testlogs/. Also did some experimentation with various documents and classification based on core number has been reasonably accurate (relevance of unsupervised core number based classification to the document's actual class decreases with core number) and top percentage classes reflect the subject matter of the document. Also best_matching_synset() lesk disambiguation implemented in AsFer and NLTK lesk() functions are faster than PyWSD simple_lesk().
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Fixed a major bug in RecursiveGlossOverlap Classifier in definition graph construction : Edges for all lemma names of previous level synsets were added which bloated the graph and somewhat skewed core number of the graph. Changed it to create edge only for one lemma name. All other Recursive Gloss Overlap implementations in InterviewAlgorithm/ folder already have this fix. Regenerated the hypergraph for shoppingcart3 example Logs for this have been added to testlogs/
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
----------------------------------------------------------------------------------------- Reinforcement Learning - ThoughtNet based Recommender Systems Implementation Update ----------------------------------------------------------------------------------------- *) Recursive Gloss Overlap classifier disambiguation has been updated to invoke NLTK lesk algorithm function() with options to choose between PyWSD lesk implementation and best_matching_synset() primitive disambiguation native implementation *) New usecase shopping cart example with book reviews has been added
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
---------------------------------------------------------------------------------------------------------------- An experimental special case k-CNF SAT Solver algorithm implementation ---------------------------------------------------------------------------------------------------------------- Algorithm: ---------- *) Create a k-DNF from k-CNF input by negating literals in all clauses of k-CNF *) Binary encode the clauses of k-DNF into set of binary strings of length n where n is number of variables *) Compute the complement of the previous binary encoded set from set of 2^n binary strings *) These are satisfying assignments *) Presently requires all literals in each CNF clause *) Implemented for simulation of a Majority Voting with Voter SATs
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
------------------------------------------------------------------------------------------------------------- Berlekamp-Welch Algorithm implementation - update for text message decoding ------------------------------------------------------------------------------------------------------------- *) Using unicode values for ordinals with ord() and inverting with chr() causes overflow errors in SciPy/NumPy Linear Algebra Equation solver (solve()) because evaluating polynomials for large messages creates huge numbers *) Hence unicode has been replaced with a simple dictionaries - alphanumeric to numeric values and numeric to alphanumeric values. *) With above 2 dictionaries an example text message with error (garbled version of "thisissentence") is list decoded with a window of possible values for each letter positions. *) Logs show an approximate decoding of original message from message with error. *) Garbled text with error was created from a previous execution of Berlekamp-Welch algorithm.
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
------------------------------------------------------------------------------------------------------------------------ Berlekamp-Welch Polynomial Reconstruction Decoder Implementation (for Reed-Solomon codes, texts, numerical points etc.,) ------------------------------------------------------------------------------------------------------------------------ *) Implements Berlekamp-Welch Polynomial Reconstruction Algorithm for reconstructing data from errors. *) System of linear equations has been solved with NumPy/SciPy Linear Algebra solve(). *) Logs show how well the reconstructed polynomial closely approximates the original polynomial. *) Error locator polynomial E is degree 1 and Numerator Q is of degree = number of points thus yielding Q/E=P (reconstructed) *) Special case is natural language text with errors and a polynomial can be reconstructed from ordinal values of unicode
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
*) Updated debug prints and logs *) Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
-------------------------------------------------------------------------------------------------------- Polynomial Encoding of a Text and 2 new distance measures based on it -------------------------------------------------------------------------------------------------------- *) It has been earlier mentioned that an alphanumeric text can be construed as a polynomial defined as f:position->alphabet *) This python implementation encodes a text as a polynomial by applying NumPy Polyfit function on ordinal values of the text letter positions as y-axis and letter positions as x-axis. *) A sample polynomial plotted with matplotlib has been added to testlogs/ alongwith logs *) Once a text is plotted as a polynomial curve, it is natural to define distance between two strings as distance between two respective polynomial encodings. *) Usual distance measures for strings e.g Edit distance are somewhat qualitative and do not quantify in numeric terms exactly. *) Distance between polynomials of two texts is numerically quite sensitive to small changes in texts and visually match the definition of "distance" between two polynomials of strings.
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
*) An example software analytics code based on DeepLearning implementation has been added to software_analytics which import BackPropagation,Convolution and RecurrentLSTM to learn models from software analytics input variables (CPU%, Memory% and TimeDuration%) *) Logs for all 3 models learnt with same input variables have been added to software_analytics/testlogs
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
AsFer-VIRGO boost::python C++ invocation works on 4.1.5 kernel and i915 related crashes were not found in kern.log.
Srinivasan Kannan
More... almost 10 years ago
Updated Code comments
Srinivasan Kannan
More... almost 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... almost 10 years ago
--------------------------------------------------------------------------------------------- DeepLearning BackPropagation Implementation Update --------------------------------------------------------------------------------------------- *) DeepLearning BackPropagation code has been changed to have 3 inputs, 3 hidden and 3 output layers with 3*3=9 input-hidden weights and 3*3=9 hidden-output weights with total of 18 weights *) Software Analytics example has been updated with a third input and logs for it have been committed to testlogs/. *) Logs include a diff notes of how to extend this for arbitrary inputs and accuracy of backpropagation which beautifully converges at ~10^-26 error after ~1000000 iterations.
Srinivasan Kannan
More... almost 10 years ago