Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 2 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
Logs for DiscreteHyperbolicFactorizationUpperbound_Bitonic_Spark.py with bidict
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Imported BiDirectional Map - bidict - for reverse lookup (for a value, lookup a key in addition to for a key, lookup a value) which is required to avoid the looping through of all keys to match against a tile element.
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Compare step in Bitonic Sort for merged tiles has been Spark MapReduced with this separate script. Can be used only in a large cluster. With this comparators are highly parallelizable over a cloud and thus in NC.
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Implementation for NC Discrete Hyperbolic Factorization with Bitonic Sort alternative (in lieu of unavailable PRAM implementations): ----------------------------------------------------------------------------------------------------------------------------------- - Bitonic Sort is O(logn) and theoretically in NC - Bitonic Sort requires O(n^2logn) parallel comparators. Presently this parallelism is limited to parallel exchange of variables in bitonic sequence for which python has builtin support (bitonic_compare()) - internally how it performs on multicore archs is not known. - a shell script has been added which does the following (cpp-src/miscellaneous/DiscreteHyperbolicFactorizationUpperbound_Bitonic.sh): - NC Factorization implementation with Bitonic Sort is C+++Python (C++ and Python) - C++ side creates an unsorted mergedtiles array for complete tesselated hyperbolic arc (cpp-src/miscellaneous/DiscreteHyperbolicFactorizationUpperbound_Bitonic.cpp) - This mergedtiles is captured via output redirection and awk in a text file cpp-src/miscellaneous/DiscreteHyperbolicFactorizationUpperbound_Bitonic.mergedtiles - Python bitonic sort implementation which is a modified version of algorithm in https://en.wikipedia.org/wiki/Bitonic_sorter requires the size of the array to be sorted in exponents of 2. Presently it is hardcoded as array of size 16384 - 2^16 (python-src/DiscreteHyperbolicFactorizationUpperbound_Bitonic.py). It outputs the sorted array. Vacant elements in array are zero-filled
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
- Updated AsFer Design Document - IharaIdentity logs committed in python-src/testlogs/complement.IharaIdentityLogs.10April2016
Srinivasan Kannan
More... over 10 years ago
Complement Function script updated with a new function to print Ihara Identity zero imaginary parts mentioned in Complement Function extended draft: 1. https://sourceforge.net/p/asfer/code/HEAD/tree/asfer-docs/AstroInferDesign.txt 2. https://github.com/shrinivaasanka/asfer-github-code/blob/master/asfer-docs/AstroInferDesign.txt
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
- Updated AsFer Design Document - KKL inequality and Denoisification - Spark-Cython intrinsic merit computation has been md5hash enabled again since there were key errors with large strings - logs and screenshots have been committed to testlogs/ - rebuilt Cython .c, .so files - Spidered text has been changed - This took 28 minutes for ~40+ keywords better than 6 minutes for ~7 keywords earlier. Again Spark in single node cluster dualcore has some bottleneck.
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
----------------------------------------------------------------------------------- Loopless Map function for Spark_MapReduce() ----------------------------------------------------------------------------------- - new Map function MapFunction2() has been defined which doesn't loop through all keywords per level, but does the glossification per keyword which is aggregated by Reduce step. MapFunction() does for loop for all keywords per level. - a dictionary cacheing with graphcachelocal has been done for lookup of already gloss-tokenized keywords. This couldn't be memcached due to a weird error that resets the graphcachelocal to a null which shouldn't have. - This has a execution time of 6 minutes overall for all tasks compared to for-loop Map function in Spark_MapReduce(). - logs and screenshots for Loopless Map with and without graphcachelocal have been committed - Rebuit Cython .c, .o, .so, intrinsic merit log and DOT files
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
-------------------------------------------------------------------------------------------------------- Spark-Cython Intrinsic Merit computation with spark-defaults.conf enabled -------------------------------------------------------------------------------------------------------- - spark-defaults2.conf has been enabled with extra java options for performance (reference: http://docs.oracle.com/javase/8/docs/technotes/tools/unix/java.html) - lookupCache and lookupCacheParents boolean flags have been done away with because memcache has been enabled by default for subgraph cacheing - With this spark logs show more concurrency and less latency. Dynamic allocation was in conflict with number of executors and executors has been hence commented in spark-defaults2.conf - text document has been updated to have ~10 keywords. This took almost 7 minutes which looks better than one with spark-defaults.conf disabled - Screenshot and logs have been committed in testlogs/
Srinivasan Kannan
More... over 10 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
--------------------------------------------------------------------------- MemCache Integration with Spark-Cython Intrinsic Merit Computation --------------------------------------------------------------------------- - In this commit, native python dictionary based cache lookup for Spark_MapReduce() and Spark_MapReduce_Parents() is replaced with standard MemCache Distributed Object Cacheing get()/set() invocations. - Logs and Screenshots for above have been committed to testlogs/ - Prerequisites are MemCached (ubuntu apt package) and Python-memcache (pip install) libraries - a recrawled webpage has been merit-computed - MemCached is better than native dict(s) because of standalone nature of Cache Listener which can be looked up from anywhere on cloud
Srinivasan Kannan
More... over 10 years ago