Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 16 hours ago. based on code collected 1 day ago.
Nov 06, 2025 — Nov 06, 2026
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
Updated AsFer Design Document
Srinivasan Kannan
More... over 11 years ago
Updated ProductOwnerProfile
Srinivasan Kannan
More... over 11 years ago
Replica of http://sourceforge.net/p/asfer/ in GitHub
Srinivasan Kannan
More... over 11 years ago
--------------------------------------------------------------------------------------------------------------- DeepLearning - Reinforcement Learning - Recommender Systems - Recursive Lambda Function Growth Merit --------------------------------------------------------------------------------------------------------------- 1.RecommenderSystems/Create_RecommenderSystems_Hypergraph.py has been changed similar to ThoughtNet/ to rank the hypergraph edges by both Sentiment and Graph Tensor Neuron Network Intrinsic Merit 2.RecommenderSystems/RecommenderSystems_Hypergraph_Generated.shoppingcart3.txt has been recreated and rankings for hyperedges through hypergraph stack vertices is different from sentiment scoring 3.DeepLearning_ReinforcementLearningRecommenderSystems.py has been changed to remove creamy layer computation for classes and all classes are lookedup in hypergraph 4.Logs for these have been committed to RecommenderSystems/testlogs/Create_RecommenderSystems_Hypergraph.GraphTensorNeuronNetworkMerit.log.12July2018 and testlogs/DeepLearning_ReinforcementLearningRecommenderSystems.shoppingcart3.log.12July2018 5.Example input is a book on quantum physics and is lookedup in Recommender Systems Hypergraph and matches return lot of books on quantum mechanics and physics
Srinivasan Kannan
More... almost 39 years ago
------------------------------------------------------------------------------------------------------------- SAT Solver Update - Randomized Rounding and other linear system of equation solvers ------------------------------------------------------------------------------------------------------------- 1.SAT Solver has been updated to include a special if condition to check if Alpha=1 and directly invoke linear system of equations solvers instead of least squares regression solvers. Various system of equations solvers like minres,cg,solve and their variants in ScipPy were tried for number_of_clauses==number_of_variables and minres was found to be the best equalling lsmr's accuracy. But they have been commented and lsmr() is invoked with no regularization damp for Alpha=1 2.Randomized Rounding of the fractional real assignments computed from least squares has been refined. Earlier the rounding threshold was set to 0.5 which has been generalized to the midpoint of the assignment array ((min + max)/2) 3. This midpoint computation is important because it finds the halfway inflexion point of the polynomial traversing the real assignment points. This polynomial is approximated to a step function of 1s and 0s based on whether the assignment for a variable is above or below half-way mark. 4. Polynomial through the fractional assignment points is sinusoidal and it crosses this midpoint in as many places as there are 0-to-1 and 1-to-0 inflexions. 5. Another randomized rounding scheme is also implemented by computing a pseudorandom fraction between 0 and 1 as threshold. Midpoint threshold is quite accourate and is mostly 0.5 which is on the expected lines, while values other than 0.5 are caused because of negative values to some variables in least squares. This follows the usual randomized rounding convention (e.g for Set Cover). 6. Few Random SAT instances of 1000 variables and Alpha=1 have been computed and MAXSAT approximation ratio has been found to be 97-98% which is captured in logs. In some other executions, this ratio touched even 98.5% 7. Accuracy of randomized rounding threshold is crucial to maximizing MAXSAT ratio. 8. Quantum HHL algorithm solves system of equations Ax=B in logarithmic time while previous classical solvers are polynomial in clauses-variables. Theoretically this SAT Solver invoking HHL on a quantum computer should be able to solve SAT NP-Complete problem approximately in logtime and thus in BQ-P (or BQ-NC).
Srinivasan Kannan
More... almost 39 years ago