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
Theory - Least Squares SAT Solver - Random Restrictions - Rounding - References
Srinivasan Kannan
More... about 8 years ago
Theory - KRW Conjecturex, Majority+VoterSAT composition, Depth Hierarchy, Communication Complexity
Srinivasan Kannan
More... about 8 years ago
Theory - Tournament Trees - Kislitsyn Theorem
Srinivasan Kannan
More... about 8 years ago
Theory - Parallel Sorting - k-mergesort for factoring - References
Srinivasan Kannan
More... about 8 years ago
Feature - Video Summary - Bugfixes
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ 612. (FEATURE) Video EventNet Core Topological Sort Summary - Bugfixes - 28 October 2018 ------------------------------------------------------------------------------------------------------------ (*) Some errors in inverse distance merit computation of the Video EventNet Tensor have been corrected which was earlier a 2-dimensional tensor and has been made 3-dimensional in this commit. (*) Core Summary of the Video has also been changed to compute maximum merit for each tensor product inverse distance merit vector and apply a threshold filter for the distance merit. (*) Three dimensional EventNet Tensor is mapped to NetworkX Graph after previous threshold filter. (*) Only causal frames above inverse distance threshold are included in the NetworkX graph and main core is computed on this graph (*) self loops and parallel edges are removed and topological sorting is performed on main core. (*) logs for this have been committed to testlogs/ImageGraph_Keras_Theano.log.28October2018
Srinivasan Kannan
More... about 8 years ago
Points-to Analysis - Reference updated
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - Video Summary by Core Topological Sort, Empath hashmap
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ 611. (THEORY and FEATURE) Video Emotions Hashmap, Video Core Topological Sort Summary - 25 October 2018 - related to 611 ------------------------------------------------------------------------------------------------------------ (*) Empath emotions of the Video EventNet Tensor Products Graph have been aggregated in a hashmap. This facilitates quantitative emotional scoring of the video. Hashmap correlates the empath emotion and its cumulative score. (*) Core (Main core) of the Video EventNet Tensor Product Graph has been computed and is sorted topologically for creating some ordering for the summary. (*) New example MP4 video has been created from 6 JPEG photos (in which I am present) and google search related to self (ExampleVideo_3.mp4) : Frame 1 - Self - Mahabalipuram, Chennai - 2012 (Copyright: Self) Frame 2 - Self - photo (Copyright: Self) Frame 3 - Self - VISA photo (Copyright: Self) Frame 4 - Self - Family photo (Copyright: Self) Frame 5 - Self - IEC Sun Microsystems - Bengaluru - 2004 (Copyright: Ex-Colleagues at Sun Microsystems) Frame 6 - Self - IEC Sun Microsystems - Bengaluru - 2000 (Copyright: Sun Microsystems/Oracle) Frame 7 and 8 - Google Search of Self (*) Previous recording is a mix of photos of varied genre. Sentiment hashmap for this video prints (excerpts): Sentiment Analysis of the Video: [('work', 1484.0), ('toy', 424.0), ('tool', 848.0), ('technology', 3180.0), ('social_media', 1060.0), ('shopping', 424.0), ('science', 1060.0), ('restaurant', 848.0), ('reading', 424.0), ('programming', 2544.0), ('play', 424.0), ('phone', 1696.0), ('office', 1060.0), ('negative_emotion', 0.0), ('musical', 424.0), ('music', 424.0), ('messaging', 1272.0), ('meeting', 424.0), ('internet', 2968.0), ('hiking', 424.0), ('furniture', 212.0), ('dance', 212.0), ('computer', 5088.0), ...]. This is a remarkable inference of the Video content by Keras-Theano ImageNet giving high weightage to office and technology as emotional annotations. (*) Logs for this have been committed to testlogs/ImageGraph_Keras_Theano.log.25October2018
Srinivasan Kannan
More... about 8 years ago
Theory - Video Summary and Topological Sort
Srinivasan Kannan
More... about 8 years ago
Theory - NC-PRAM equivalence - reference
Srinivasan Kannan
More... about 8 years ago
Theory - Video Summarization by EventNet Dense Subgraphs
Srinivasan Kannan
More... about 8 years ago
Theory - EventNet Tensor Products - Example for Movie and Youtube Merit
Srinivasan Kannan
More... about 8 years ago
Theory - Gomory Cut - Reference
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - EventNet Tensor Products Merit of Large Scale Visuals - Emotional Merit
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ 608. (THEORY and FEATURE) Video EventNet Tensor Products - Emotional Merit of Large Scale Visuals - 15 October 2018 ------------------------------------------------------------------------------------------------------------ 1. EventNet Tensor Product representation of a Large Scale Visual just infers the connectedness of the information in the video with no emotional or sentiment scoring. 2. To redress this, new function has been included in ImageGraph_Keras_Theano.py to lookup the edges of tensor product graph in empath and aggregate the sentiments of the vertices as a tuple which is qualitative emotive merit of the video. 3. Two new example visuals have been added: ExampleImage_1.jpg and ExampleVideo_2.mp4 in addition to earlier ExampleVideo_1.mp4 and images. 4. Keras has mistook ExampleImage_1.jpg to be a theatre audience which is indeed a classroom photo and annotated it accordingly (probably because seating and ambience are similar in both). Also Empath has misjudged the alumni website content in ExampleVideo_2.mp4 to large extent excluding "college" as emotional annotation. Logs have been committed to testlogs/ImageGraph_Keras_Theano.log.15October2018 5. These Photo and Videos are courtesy (and related to self): (*) ExampleVideo_1.mp4 - https://www.cmi.ac.in/people/fac-profile.php?id=shrinivas - Chennai Mathematical Institute - Research Scholar website - 2010-11 (*) ExampleImage_1.jpg - http://www.angelfire.com/id/95cse/album.html - PSG College of Technology - 1995-99 CSE batch alumni website album - Classroom Photo (class1.jpg) (*) ExampleVideo_2.mp4 - https://alumni.psgtech.ac.in/profile/view/srinivasan-kannan-1 - PSG College of Technology Official Alumni website 6. Frames captured by OpenCV2 are prefixed by the Video file name. 7. THEANO_FLAGS have been set as: declare -x THEANO_FLAGS="cxx=/usr/bin/g++-6"
Srinivasan Kannan
More... about 8 years ago
Theory - Randomized rounding - cutting planes - Gomory cut for rounding real feasible solutions to integer solutions in ILPs - reference
Srinivasan Kannan
More... about 8 years ago
Theory - Orthogonal Range Reporting for Factorization - Reference
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - Video EventNet Tensor Products - Inverse Distance Merit of Large Scale Visuals
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------ Video EventNet Tensor Products - Inverse Distance Merit of Large Scale Visuals ------------------------------------------------------------------------------------------------------------ 1.New function inverse_distance_intrinsic_merit() has been included in ImageGraph_Keras_Theano.py. It iterates through each tensor product entry of Video EventNet Tensor Products matrix and for every edge (a,b)-(c,d) in tensor product computes inverse wordnet distance (which is basis for ImageNet): 1/dist(a,b)*1/dist(c,d) and populates a weights matrix per tensor product graph. 2.Each tensor product in EventNet Tensor corresponds to causality between any 2 frames in the video. 3.An example merit computation log has been committed to testlogs/ImageGraph_Keras_Theano.log.4October2018 which shows merits for 4 tensor products (2 frames * 2 frames). 4.Lot of weights in merit matrix in logs have huge values implying high causality between two frame events. 5.EventNet Tensor Product Algorithm is computationally intensive. Each ImageGraph per frame encodes relationship between objects (actors) in the frame event. 6.Size of the EventNet Tensor is O(number_of_frames * number_of_frames * average_number_of_actors_per_frame * average_number_of_actors_per_frame) and memory intensive too.
Srinivasan Kannan
More... about 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - VideoGraph EventNet Tensor Products Merit for Large Scale Visuals
Srinivasan Kannan
More... about 8 years ago
-------------------------------------------------------------------------------------------------------------- 606. (THEORY and FEATURE) VideoGraph EventNet Tensor Product Merit for Large Scale Visuals - 3 October 2018 -------------------------------------------------------------------------------------------------------------- 1. Two new functions have been committed to ImageGraph_Keras_Theano.py for reading a Video file, extracting frames from the video upto certain number and extract object information by Keras-Theano ImageNet predictions for each frame. 2. imagenet_videograph() function depends on OpenCV 3.4.3 (cv2) Python for reading Videos and writing Frame JPEG files suffixed by unique id. Each frame is mapped to a TextGraph - ImageGraph - by imagenet_imagegraph() 3. videograph_eventnet_tensor_product() function computes tensor products of pairs of Frame ImageGraphs and returns an EventNet Tensor from NetworkX. 4. logs, an example Video (MP4) and 2 sample extracted frames (JPEG) for EventNet Tensor representation of a video have been committed to testlogs/ImageGraph_Keras_Theano.log.3October2018 5. From this EventNet Tensor any kind of intrinsic merit can be computed - e.g previously described ImageNet based inverse weights, Graph Edit Distance between two Frame ImageGraphs, Tensor Rank for connectedness etc.,
Srinivasan Kannan
More... about 8 years ago
Theory - Sequential Optimization for Factorization - Ray Shooting and Planar Point Location
Srinivasan Kannan
More... about 8 years ago
Correction - Ray Shooting and Planar Point Location - Error estimate
Srinivasan Kannan
More... about 8 years ago
Theory - Parallel Planar Point Location, Ray Shooting Queries, Multisearch in the context of factorization
Srinivasan Kannan
More... about 8 years ago
Feature - Music Pattern Mining - Notes detection
Srinivasan Kannan
More... about 8 years ago
------------------------------------------------------------------------------------------------------------- 604. (FEATURE) Music Pattern Mining - Onset of Notes Detection - 1 October 2018 - related to 67,68,69 ------------------------------------------------------------------------------------------------------------- 1. audio_features() function in AudioToBitMatrix.py has been updated to compute notes onset from audio signal 2. Notes Onset Detection involves finding peaks in onset strength of the signal time series. 3. Extracting notes from signal converts the music time series to a string from a language of notes alphabets. 4. Most algorithms for audio analysis are time-series (fft,statistics etc.,) based. Considering music as a Formal language opens new vistas into theoretical analysis of music. For example, finding a deterministic finite state automaton for sequence of notes in a music signal is measure of complexity in music. 5. Not just automata, variety of string complexity measures (e.g Kolmogorov complexity - MDL) can effectively describe music notes strings.
Srinivasan Kannan
More... about 8 years ago
Updated Requirements.txt
Srinivasan Kannan
More... about 8 years ago
Theory and Feature - ImageGraph implemented for Large Scale Visuals - Keras-Theano
Srinivasan Kannan
More... about 8 years ago