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
--------------------------------------------------------------------------------------------------------------- 627. (FEATURE) Vowelless Text (De)Compression - simplified likelydict MLE for HMM - 3 January 2019 --------------------------------------------------------------------------------------------------------------- 1. TextCompression.py has been classified to new class VowellessText and all functions have been made its members. 2. One time Likelyhood Dictionary computation of prefixes,suffixes and substrings has been shifted to WordSubstringProbabilities.py - wordlikelydict() and wordprefixsuffixsubstringsprobdist() 3. All precomputed prefix and suffix priors are initialized in TextCompression.py __init__() 4. Costly loop in HiddenMarkovModel_MLE() of TextCompression.py has been done away with and instead wordlikelydict() has been changed to choose between costly exhaustive search and substringsdict list comprehension. 5. Hardcoded priors for HMM have been removed and replaced by previous likelihood computation. 6. Logs for this have been committed to testlogs/TextCompression.log.3January2019 7. likelydict is created on demand once per compressed word than earlier heavy initialization for all possible likelihoods. 8. Regular Expressions are used in list comprehension for matching strings - "_" are treated as alphabets in lower or upper case. 9. For Exhaustive Search. following line is crucial which computes the conditional probability of a prefix for a suffix: likelydict[k3[:-1]+k3[len(k3)-1]+k4[1:]] = v3 * v4 10. The else clause, generalizes it and applies precomputed frequency probabilities of concatenated substrings circumventing the following conditional probability: likelydict[substring1 + substring2 + ... + substringN] = Pr(substring1) * Pr(substring2) * ... * Pr(substringN) by: likelydict[substring1 + substring2 + ... + substringN] = Pr(substring1 + substring2 + ... + substringN) 11. Both should be almost equal if substring random variables are independent and identically distributed. 12. An example English text from https://www.valmikiramayan.net/utf8/yuddha/sarga128/yuddha_128_frame.htm - Verse 120 has been compressed and decompressed. 13. Decompressed text has almost more than 50% accuracy and most priors are equal misleading the decipherment. 14. Previous HMM based probabilistic decompression can be compared to List decoding Berlekamp-Welch implementation which maps text to a polynomial
Srinivasan Kannan
More... almost 8 years ago
Theory - Section 555 updated - PH^A and BQP^A, Margulis-Russo, Condorcet Jury Theorem
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines in Hilbert Spaces, Linear Transformations between Hilbert Spaces, Logical Formulas embedded in Hilbert Spaces
Srinivasan Kannan
More... almost 8 years ago
References - Triadic Closure
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines, Halting Problem, Turing Degrees
Srinivasan Kannan
More... almost 8 years ago
References - Diophantine Sets and Turing Degrees
Srinivasan Kannan
More... almost 8 years ago
References - Shell Turing Machines and Physics
Srinivasan Kannan
More... almost 8 years ago
Theory - Turing Degrees, Turing Machines defined over Hilbert Spaces, Application to Quantum Computation - References
Srinivasan Kannan
More... almost 8 years ago
Theory - Partial Ordered Intrinsic Merit Rankings, Galois Connections, ThoughtNet
Srinivasan Kannan
More... almost 8 years ago
Theory - References - Embedding Turing Machines in Hilbert Space
Srinivasan Kannan
More... almost 8 years ago
Theory - Undecidability of Halting Problem and Shell Turing Machines
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines, Turing Degrees
Srinivasan Kannan
More... almost 8 years ago
Theory - People Analytics, Majority Function, Margulis-Russo Threshold, Feasibility of Automatic Crowdsourced Recruiting, Intrinsic Merit, Fame
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machine - Example sentence
Srinivasan Kannan
More... almost 8 years ago
Theory - Turing Jump and Shell Turing Machines embedded in vector space
Srinivasan Kannan
More... almost 8 years ago
References - Computational Geometric Factorization by AKS Sorting Networks in LogDepth - O(logN) time sorting network exists
Srinivasan Kannan
More... almost 8 years ago
Reference - Sorting Networks
Srinivasan Kannan
More... almost 8 years ago
References - Social Networks - Viruses and Fads
Srinivasan Kannan
More... almost 8 years ago
Recursive Lambda Function Growth - TextGraph Machine Translation
Srinivasan Kannan
More... almost 8 years ago
------------------------------------------------------------------------------------------------------------- 625. (THEORY) Recursive Lambda Function Growth - TextGraph Machine Translation - 10 December 2018 - related to 178 ------------------------------------------------------------------------------------------------------------- 1. RecursiveLambdaFunctionGrowth python implementation has a new function to do machine translation of Text Graph english word vertices to another natural language. It uses PyDictionary Google Translate REST API service internally. 2. An example Text Graph from Recursive Lambda Function Growth has been translated from English to Kannada (an Indian language). 3. This is just a primitive implementation without sentences creation which can be done through usual graph traversal or random walks. 4. Logs have been committed to testlogs/RecursiveLambdaFunctionGrowth.log.MachineTranslation.10December2018
Srinivasan Kannan
More... almost 8 years ago
Theory - Shell Turing Machines, MRDP Theorem, Turing Degrees, Embedding Formal Languages in Vector Space
Srinivasan Kannan
More... almost 8 years ago
Theory - Arithmetization and Diophantine representation of 3SAT
Srinivasan Kannan
More... almost 8 years ago
Theory - Arithmetic Circuit Complexity and Diophantine Complexity, Universal Diophantine Equation, ABC Conjecture
Srinivasan Kannan
More... almost 8 years ago
Feature - Vowelless text decompression - HMM prefix and suffix priors from English Dictionary
Srinivasan Kannan
More... almost 8 years ago
--------------------------------------------------------------------------------------------------------------- 622. (THEORY and FEATURE) Text (De)Compression by HMM on Vowelless texts - Prefix and Suffix Probabilities from English Dictionary - 6 December 2018 --------------------------------------------------------------------------------------------------------------- 1.Python implementation for Decompressing vowelless compressed text TextCompression.py has been updated to invoke Prefixes and Suffixes probability priors computed from English Dictionary instead of hardcoded priors. 2.New source file WordSubstringProbabilities.py for computing word prefixes and suffixes probabilities by FreqDist has been committed which parses English Wordlist from Dictionary.txt into two probability distribution dictionaries for prefixes and suffixes. 3.Function wordprefixsuffixprobdist() in WordSubstringProbabilities.py called in TextCompression.py has been nominal because of computationally intensive Maximum Likelyhood Estimator in Hidden Markov Model (likelydict) and hardcoded priors are used. 4.Logs for this commit have been committed to WordSubstringProbabilities.log.6December2018 and TextCompression.log.6December2018.gz 5.compressedtext.txt and decompressedtext.txt have been updated 6.MLE likelydict is O(prefixes*suffixes) = O(number_of_words^2*length_of_longest_word^2) is costly one time disk read and might require some memcache-ing during initialization.
Srinivasan Kannan
More... almost 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... almost 8 years ago
Theory - Chaos and Diophantines
Srinivasan Kannan
More... almost 8 years ago
Theory - References - Complement Diophantines - Beatty Sequences - Complementary Sets - ABC Conjecture
Srinivasan Kannan
More... almost 8 years ago
Feature - Dictionary filter for names - People Analytics
Srinivasan Kannan
More... almost 8 years ago
--------------------------------------------------------------------------------------------------------------- 621. (FEATURE) Social Network Analysis - People Analytics - Dictionary filter for names - 4 December 2018 --------------------------------------------------------------------------------------------------------------- 1.RecursiveGlossOverlap_Classifier.py has been changed to define a new function nondictionaryword() for filtering nondictionary words. This is necessary for parsing text of Social Profiles which contain names in SocialNetworkAnalysis_PeopleAnalytics.py. Finding names in text is an open AI problem which comes under the purview of Named Entity Recognition (NER) which often requires CRF or Neural Network Training models. 2.As an alternative to statistical learning, if a word in text is not in a language Dictionary it is assumed to have exited any semantic network (WordNet for English etc.,) which could be names of people, places etc., But there are some exceptions to this and countably few names are in language dictionaries and encyclopedia. 3.Function nondictionaryword() looksup a word in PyDictionary which is based on WordNet, Google Translation, Thesaurus.com and returns true or false based on language dictionary. If for a word lookup null meaning is returned, it is outside Semantic Network and nondictionaryword() returns true. 4.Depth of the Recursive Gloss Overlap recursion has been parametrized in RecursiveGlossOverlap_Classifier.py and RecursiveLambdaFunctionGrowth.py and invoked in SocialNetworkAnalysis_PeopleAnalytics.py for depth 2. 5.WordNet and ConceptNet consist of network of concepts than textual words. PyDictionary has additional support for other datasources - Google Translation and Thesaurus - specific to language words.
Srinivasan Kannan
More... almost 8 years ago