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 01, 2025 — Nov 01, 2026
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
-------------------------------------------------------------------------------------- Code Optimizations - Cacheing and Loop invariant - Interview Algorithm -------------------------------------------------------------------------------------- - New Cacheing datastructures have been included for storing already found previous level backedge vertices for each level and already found gloss tokens. Presently these caches are python dictionaries only and in a distributed setting this can be replaced with an object key-value stores like memcached.(InterviewAlgorithm/InterviewAlgorithmWithIntrinisicMerit_SparkMapReducer.py) - There was an inefficient loop invariant computation for prevelevesynsets_tokens which has been moved to outermost while.(InterviewAlgorithm/InterviewAlgorithmWithIntrinisicMerit_Crawl_Visual_Spark.py) - spark-default2.conf has been updated with few more Spark config settings - Spidered text has been recrawled - logs and screenshots for this commits have been included in testlogs/ - Cython .c , .so files have been rebuilt - Cacheing above in someway accomplishes the isomorphic node removal by way of Cache lookup
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
--------------------------------------------------------- More benchmarking with new spark-defaults.conf settings: --------------------------------------------------------- - an enlarged spark-defaults.conf has been committed as spark-defaults2.conf - Performance slowdown with spark-defaults.conf remains same with almost 4 times slowdown - logs and screenshots have been committed in InterviewAlgorithm/testlogs/
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
More benchmarking of PySpark-Cython Intrinsic Merit computation ---------------------------------------------------------------- - Enabled Spark RDD cacheing with cache() - storage level MEMORY - Recompiled .c and .so with Cython - uncommented both lines in Cython setup.py - logs and screenshots for above have been committed in testlogs/ - locking.acquire() and locking.release() have been commented - With this per task duration has been brought down to ~402 milliseconds on single node cluster. Ideally on a multinode cluster when tasks are perfectly distributable, this should be the runtime.
Srinivasan Kannan
More... over 10 years ago
- Updated AsFer Design Document - Scrapy crawl spidered HTML webpage text
Srinivasan Kannan
More... over 10 years ago
PySpark-Cython Interview Algorithm for Merit - Benchmark -------------------------------------------------------- - 3 executions with and without spark-defaults.conf config settings have been benchmarked - Cython setup.py has been updated to compile both InterviewAlgorithmWithIntrinisicMerit_Crawl_Visual_Spark.py and InterviewAlgorithmWithIntrinisicMerit_SparkMapReducer.py to create .c and .so files - Benchmark yields some intriguing results: - Executions with spark-defaults.conf enabled are 4 times slower than executions without spark-defaults.conf - Execution without spark config is twice faster than earlier benchmark - Logs and screenshots for above have been committed to testlogs - Text document has been recrawled and updated (number of keywords almost same as previous benchmark) - Spark config file has been committed
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 Requirements.txt
Srinivasan Kannan
More... over 10 years ago
renamed: python-src/testlogs/DeepLearning_ConvolutionNetwork.out2.15March2016 -> python-src/testlogs/DeepLearning_Backpropagation.out2.15March2016
Srinivasan Kannan
More... over 10 years ago
Increased number of backpropagation iterations to 3000000. Logs for last few iterations have been committed in testlogs/. But error tapering is quite slow. Might require few million iterations more.
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
- DeepLearning BackPropagation implementation: - An example Software Analytics usecase for CPU and Memory usage has been included - Number of Backpropagation weight update iterations has been increased 10 fold to 1000000. - logs for some iterations have been included in testlogs/ - DeepLearning Convolution Network implementation: - Image patterns have been changed to 0, 8 and no-pattern. - Final neuron layer weights have been changed by squaring to scale down the output - this differentiates patterns better. - logs for this have been included in testlogs/
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document - Section 239
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
- Some bugs resolved - Added one more example with no pattern - Convolution is computed for all 3 bitmap examples - Final neuron layer now is a function of each point in all maxpooling layers - The existence of pattern is identified by the final output of each of 10 neurons - Patterns 0 and 3 have a greater neural value than no pattern. Gradation of neural value indicates intensity of pattern. - Above is a very fundamental pattern recognition for 2 dimensional data. Sophisticated deconvolution is explained in http://arxiv.org/abs/1311.2901 which reverse engineers pooling layers with Unpooling. - 3 logs for this commit have been included in testlogs/ - random weighting has been removed.
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
========================================================================= Deep Learning Convolution - Multi Feature Convolution Map Kernel Filters ========================================================================= - Convolution Network supports Multiple Feature Maps (presently 3) - New example bitmap for feature recognition of pattern "3" inscribed as 1s has been introduced - Final neural network from Max pooling layer now is randomized with randint() based weights for each of the 10 neurons - Logs for this have been committed to testlogs - Logs show a marked swing in Maxpooling map where the segments of pattern "3" are pronounced. - Final neural layer shows a variegated decision from each neuron for corresponding 3 convolution maps
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
- Updated AsFer Design Document - Section 234 - perl build idiosyncrasies - Python wrapper to invoke perl WordNet::Similarity has been added to repository - logs for above in python-src/testlogs/
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
*) New perl-src folder with perl code for computing WordNet Distance with Ted Pedersen et al WordNet::Similarity CPAN module *) Logs for some example word distances have been included in testlogs. Some weird behaviour for similar word pairs - distance is 1 instead of 0. Probably a clockwise distance rather than anticlockwise.
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document
Srinivasan Kannan
More... over 10 years ago
Updated AsFer Design Document - Section 232
Srinivasan Kannan
More... over 10 years ago