Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 9 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) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - Block Composed Boolean Function Polynomial Approximations
Srinivasan Kannan
More... over 7 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 7 years ago
(FEATURE) People Analytics - Kaggle LinkedIn Dataset Analysis
Srinivasan Kannan
More... over 7 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 685, (FEATURE) People Analytics - Kaggle LinkedIn Dataset Analysis - 26 July 2019 -------------------------------------------------------------------------------------------------------------------------------------------- 1. This commit introduces a new function in People Analytics implementation - linkedin_dataset_tenure_analytics() - which parses Kaggle LinkedIn dataset from https://www.kaggle.com/killbot/linkedin and computes various correlation coefficients between column attributes. 2. Following are the People Analytics field attributes in LinkedIn obfuscated dataset which has almost 15000 rows: ['avg_n_pos_per_prev_tenure', 'avg_pos_len', 'avg_prev_tenure_len', 'c_name', 'm_urn', 'n_pos', 'n_prev_tenures', 'tenure_len', 'age', 'beauty', 'beauty_female', 'beauty_male', 'blur', 'blur_gaussian', 'blur_motion', 'emo_anger', 'emo_disgust', 'emo_fear', 'emo_happiness', 'emo_neutral', 'emo_sadness', 'emo_surprise', 'ethnicity', 'face_quality', 'gender', 'glass', 'head_pitch', 'head_roll', 'head_yaw', 'img', 'mouth_close', 'mouth_mask', 'mouth_open', 'mouth_other', 'skin_acne', 'skin_dark_circle', 'skin_health', 'skin_stain', 'smile', 'african', 'celtic_english', 'east_asian', 'european', 'greek', 'hispanic', 'jewish', 'muslim', 'nationality', 'nordic', 'south_asian', 'n_followers'] which include average variables for tenure, positions or designations per tenure, number of previous tenures, tenure lengths etc., 3.Pairwise kendall-tau Correlation coefficients of 3 important variables are computed: - average number of positions per previous tenure - average length of a position - average length of previous tenures 4.Some preprocessing of the linkedin CSV dataset was required to cast the type to integer 5.VectorAssembler is instantiated to create necessary subset projections of the linkedin dataset columns 6.5 important columns of the Spark CSV dataframe are sliced by list comprehension to 5 arrays for (ethnic, emotional, physical and racial fields are ignored): - average number of positions per previous tenure - average length of a position - average length of previous tenures - number of previous tenures - tenure length and mapped to a Pandas DataFrame which supports correlation matrix for all datatypes 7.Pandas correlation matrix is printed for these 5 fields: avg_n_pos_per_prev_tenure avg_pos_len avg_prev_tenure_len n_prev_tenures tenure_len avg_n_pos_per_prev_tenure 1.000000 0.804373 0.836112 0.836637 0.832915 avg_pos_len 0.804373 1.000000 0.983461 0.978873 0.993264 avg_prev_tenure_len 0.836112 0.983461 1.000000 0.968333 0.975659 n_prev_tenures 0.836637 0.978873 0.968333 1.000000 0.981673 tenure_len 0.832915 0.993264 0.975659 0.981673 1.000000
Srinivasan Kannan
More... over 7 years ago
(THEORY) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - Size of Majority Voting Circuit Composition and estimated runtime for Turing Machine for it
Srinivasan Kannan
More... over 7 years ago
(THEORY) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - Universality of Polynomial Threshold Functions - any boolean function can be computed by PTF - References
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Urban Planning/GIS Image Analytics - scikit learn extract_patches_2d() integration
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------- 684. (FEATURE) Urban Planning/GIS Image Analytics - scikit learn extract_patches_2d() integration - 25 July 2019 ------------------------------------------------------------------------------------------------------------------------------------------- 1. Scikit Learn extract_patches_2d() has been invoked in a wrapper function analyze_remotesensing_2d_patches() for patches analysis of Remote Sensing Imagery 2. Patch ndarrays are flattened and pixel histogram is computed for each patch of an image 3. Maximum pixel bucket in the patch histogram is an indicator of class of the patch. 4. This is in addition to RGB patch analysis function implemented earlier for Red-Green-Blue channel classification of an image 5. Logs of extract_patches_2d() and the patch histograms for 2 example remote sensing images are at ImageNet/testlogs/ImageGraph_Keras_Theano.log.RemoteSensingGIS.25July2019
Srinivasan Kannan
More... over 7 years ago
(THEORY) Proof of KRW Conjecture for Majority+VoterSAT/VoterPTF composition
Srinivasan Kannan
More... over 7 years ago
KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - MAXSAT to PTF reduction
Srinivasan Kannan
More... over 7 years ago
(THEORY) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - Reference - MAXSAT to PTF CircuitSAT
Srinivasan Kannan
More... over 7 years ago
KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity - Update
Srinivasan Kannan
More... over 7 years ago
(THEORY) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity
Srinivasan Kannan
More... over 7 years ago
Theory - KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity
Srinivasan Kannan
More... over 7 years ago
(THEORY) KRW Conjecture, Majority + VoterSAT Boolean Function Composition, BKS Conjecture, Computing any function by Parity
Srinivasan Kannan
More... over 7 years ago
(THEORY and FEATURE) Intrinsic Merit of Music - MFCC Earth Mover Distance similarity based Music Recommender
Srinivasan Kannan
More... over 7 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 681. (THEORY and FEATURE) Intrinsic Merit of Music - MFCC Earth Mover Distance similarity based Music Recommender - 12 July 2019 -------------------------------------------------------------------------------------------------------------------------------------------- 1.Mel Frequency Cepstral Coefficients of a music waveform are concise measures of music genre and have been included in AudioToBitMatrix implementation 2.Music Similarity Measure based on MFCC spectral signatures is described in http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.13.8524&rep=rep1&type=pdf which clusters music files based on earth mover distance between MFCCs of any two music files 3.Earth Mover Distance also known as Wasserstein distance is a function in SciPy - https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.wasserstein_distance.html 4.Earth Mover Distance is the amount of work required to transform one distribution (set partition or histogram) to another distribution (another set partition or histogram) and thus is distance measure between two set partitions. 5.A miniature music recommender systems has been implemented based on MFCC earth mover distance which analyzes pair of music files for MFCCs and finds the earth mover similarity between the mean of MFCCs of two music files. 6.Pairwise earth mover distances (EMD) are sorted by values of similarity (low EMD values imply high similarity) 7.3 music files have been pairwise analyzed for MFCC EMD similarity for example (2 files of Bach Sonata and 1 file of Veena Instrumental - S.Balachander's Amritavarshini - respective copyrights attributed) 8.Though number of files analyzed is small, Bach Sonata files have high EMD similarities compared to Bach and Veena Instrumental suggesting a clustering of oriental and western classical music in terms of spectral analysis.
Srinivasan Kannan
More... over 7 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Intrinsic Merit of Audio - Speech Recognition Integration and Graph Tensor Neuron Merit of the recognized speech text
Srinivasan Kannan
More... over 7 years ago
---------------------------------------------------------------------------------------------------------------------------------------------- 680. (FEATURE) Intrinsic Merit of Audio - Speech Recognition Integration and Graph Tensor Neuron Merit of the recognized speech text - 9 July 2019 ---------------------------------------------------------------------------------------------------------------------------------------------- 1.AudioToBitMatrix python implementation has been updated to include a new function which reads an audio WAV file of human speech and invokes speech recognizer (PocketSphinx) on the audio to get speech text and finds the Recursive Lambda Function Growth intrinsic merit of the speech text 2.Audio merit is split into two - Speech and Music 3.An example speech audio file (voice of author - K.Srinivasan - explains the 6 types of intrinsic merit analytics implemented in NeuronRain AstroInfer - text,video,speech,music,people-experiential-intrinsic, people-attrition or HR analytics) is used for speech recognition - music_pattern_mining/testlogs/AudioGraphExample_SpeechRecognition_2019-07-09-103018.wav 4.Prerequisite packages to be installed are Python SpeechRecognition, PocketSphinx, Swig, ALSA development library, PulseAudio development library 5.PocketSphinx recognized speech text is printed (though not accurate) and Recursive Lambda Function Growth grow_lambda_function3() is invoked on recognized speech text to create textgraph, grow lambda functions from textgraph random walks or cycles and print various merit quantities 6.This assumes Speech Recognizer is accurate to large extent.
Srinivasan Kannan
More... over 7 years ago
Theory - Majority Hardness Amplification Lemma, Noise Sensitivity, Sensitivity, Block Sensitivity - Update
Srinivasan Kannan
More... over 7 years ago
(FEATURE and THEORY) Fraud Analytics - Principal Component Analysis and KMeans++ clustering of an example Credit Card Transactions Dataset
Srinivasan Kannan
More... over 7 years ago
-------------------------------------------------------------------------------------------------------------------------------------------- 679. (FEATURE and THEORY) Fraud Analytics - Principal Component Analysis and KMeans++ clustering of an example Credit Card Transactions Dataset - 8 July 2019 -------------------------------------------------------------------------------------------------------------------------------------------- 1.This commit implements a rudimentary primitive fraud analytics Python class 2.Fraud Analytics in the context of NeuronRain has already been implemented to some extent in USBmd as wireless LAN traffic analytics and ftrace based analysis which is a low level intrusion fraud analytics 3.To broaden the purview new file FraudAnalytics.py has been added to AstroInfer repositories and a basic fraud analysis function for credit card data set from https://www.kaggle.com/mlg-ulb/creditcardfraud/downloads/creditcardfraud.zip/3 has been implemented - creditcard_fraud_analytics() 4.Traditional algorithm to analyze frauds is to find the outliers in a dataset which are deviations from the normal. 5.This function applies kmeans++ clustering from scikit learn to classify the previous kaggle creditcard transactions dataset into two clusters 0 (normal) and 1 (potential fraud). 6.Kaggle dataset is in CSV format and first 500 rows of it are read by python CSV reader and a NumPy ndarrays are created from it which splits the transactions into 150 rows training set and 350 rows test set. 7.Columns of the CSV file are Time, 28 columns which obfuscate the identity of the cardholder, Amount, label 0 or 1 (normal or fraud) which are real transactions happened in 2013 and each column is considered as a dimension for tranasaction vector 8.Classifying high dimensional data has variety of clustering algorithms including kmeans++ (https://scikit-learn.org/stable/modules/clustering.html) 9.To simiplify high dimensional data, convention followed is to apply Principal component Analysis to reduce the number of dimensions and then do clustering. e.g https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html, https://scikit-learn.org/stable/auto_examples/cluster/plot_kmeans_digits.html#sphx-glr-auto-examples-cluster-plot-kmeans-digits-py 10.Therefore previous credit card transactions vectors are clustered by fit() of the train dataset followed by predict() for test dataset. 11.Both predictions- with and without PCA - have been printed for 350 rows test dataset (0 - normal,1 - fraud): kmeans estimator pred() labels: [1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0] kmeans estimator pca pred() labels: [1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0] 12.Another alternative is to consider the dataset as multiple concurrent timeseries and plot the curves per cardholder identity. Spikes in timeseries are outliers which could be potentially fraudulent.
Srinivasan Kannan
More... over 7 years ago
(THEORY) Majority Hardness Amplification Lemma, Noise Sensitivity, Sensitivity, Block Sensitivity
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Image Analytics - ImageNet Keras-Theano - Unsupervised Recursive Gloss Overlap classifier based on ImageNet
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------- 677. (FEATURE) Image Analytics - ImageNet Keras-Theano - Unsupervised Recursive Gloss Overlap classifier based on ImageNet - 5 July 2019 ------------------------------------------------------------------------------------------------------------------------------------------- 1.As opposed to Random Forest classifier implemented earlier based on scikit learn, NeuronRain unsupervised graph theoretic Recursive Gloss Overlap classifier has been applied for some test images 2.Recursive Gloss Overlap classify() function has been invoked from imagenet_imagegraph() which was earlier only creating text graph from Keras predictions text for the image. 3.Example Dense subgraph (core number) classifications for these images in the logs at testlogs/ImageGraph_Keras_Theano.log.ImageNetClassifier.5July2019 are reasonably accurate and classify the images better. No training labelled data are required.
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Image Analytics - ImageNet Keras-Theano - Random Forest Classification wrapper for images
Srinivasan Kannan
More... over 7 years ago
------------------------------------------------------------------------------------------------------------------------------------------ 676. (FEATURE) Image Analytics - ImageNet Keras-Theano - Random Forest Classification wrapper for images - 4 July 2019 ------------------------------------------------------------------------------------------------------------------------------------------ 1.New wrapper function has been implemented which invokes scikit learn Random Forests - https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html#sklearn.ensemble.RandomForestClassifier.predict to classify set of test images based on training images and training labels. 2.This wrapper function random_forest_image_classification() pre-processes the training and test images by loading to CV2 and flattening the image ndarray and then reshaping and transposing it so that scikit learn accepts. 3.First the random forest classifier which is a forest of lot of decision trees which vote to decide the class label of a sample, is trained by fit() for training images and then test images are labelled by predict() 4.Logs for this are committed to ImageGraph_Keras_Theano.log.RandomForests.4July2019 5.Scikit learn provides train_test_split function which divides set of samples into training and test images based on which fit() and predict() are invoked. 6.Alternative version without training and test split also follows the above as in logs.
Srinivasan Kannan
More... over 7 years ago
(FEATURE) Urban Planning Analytics - Analysis of Remote Sensing GIS Imagery of Urban Sprawls
Srinivasan Kannan
More... over 7 years ago
--------------------------------------------------------------------------------------------------------------------------------------------- 675. (FEATURE) Urban Planning Analytics - Analysis of Remote Sensing GIS Imagery of Urban Sprawls - 3 July 2019 --------------------------------------------------------------------------------------------------------------------------------------------- 1.Analyzing Builtup land, Water bodies etc., of an urban sprawl is crucial for urban planners and is an apt application of image analysis. Urban sprawl is defined as expanse of urbanization in suburbs and is not measured by governance limits. For example rankings of metropolitan areas in India is: https://en.wikipedia.org/wiki/List_of_metropolitan_areas_in_India ("... Rank Metropolitan area State/Territory Population Area (in km2/sq mi) 1 Central National Capital Region Delhi, Haryana, Uttar Pradesh 25,735,000 (2016)[3] 2,163 (835)[3] 2 Mumbai Metropolitan Region Maharashtra 20,800,000 (2005)[4] 4,354 (1,681)[5] 3 Kolkata metropolitan area West Bengal 14,720,000 (2001)[6] 1,851 (715)[7] 4 Chennai metropolitan area Tamil Nadu 13,300,253 (2011) 1,189 (459)...") which is measure of urbanization.
Srinivasan Kannan
More... over 7 years ago