Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Very Low Activity

Commits : Listings

Analyzed about 3 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
Section 1440 updated
Srinivasan Kannan
More... over 3 years ago
+1440. (JIRA,THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - Concert synthesis by multiple instrument orchestra overlay on Piano music - through Music21 MIDI,MuseScore,FluidSynth,Pydub support - refinements to synthesized notes based on WFA state transition probability distribution - 28 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434,1438,1439 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks,Neo-Riemannian Theory of Music
Srinivasan Kannan
More... over 3 years ago
---------------------------------------------------------------------------------------------------------------------------------------------- 1440. (JIRA,THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - Concert synthesis by multiple instrument orchestra overlay on Piano music - through Music21 MIDI,MuseScore,FluidSynth,Pydub support - refinements to synthesized notes based on WFA state transition probability distribution - 28 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434,1438,1439 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks,Neo-Riemannian Theory of Music ---------------------------------------------------------------------------------------------------------------------------------------------- 1440.1 python-src/music_pattern_mining/AudioToBitMatrix.py has been updated to add an error condition to check length of notes sequence before Hurst exponent is computed (minimum length required is 100). 1440.2 Function percussion_synthesis(virtual_piano_notes=None,percussions=["Guitar","Trumpet","Violin","SteelDrum"],tempo=2000) has been enhanced to support an orchestra of multiple instruments including Violin,Guitar and Trumpet along with Drums in percussions parameter. Score object is instantiated in the beginning and notes composed for every instrument as Part object is inserted at 0.0 point in the Score. Score composed is written to MIDI file ./music21_percussion_for_piano.midi. 1440.3 Function music_synthesis(training_music=None,dur=5,samplerate=44100,polynomial_interpolation=True,polyfeatures=False,virtual_piano_notes=None,tempo=1,amplitude=4096,musicgenre="WesternClassical",musicwfa_notes_weights=None) has been enhanced to included music WFA weights for notes (obtained from Music WFA) as additional argument. This parameter is for a future usage. 1440.4 Function notes_to_audio(automaton=False, function=None, deterministic=True, samplerate=44100, fractal=True, periodicity=500, weightedautomatadotfile=None,state2notedict=None,genre="WesternClassical",wfaweight_threshold=0.0) has been updated to include WFA state transition weight threshold set to 0.0 by default. 1440.5 Till now the weights of WFA state transition were overlooked while computing successors which has been redressed by obtaining the weights of each of the successor and computing a probability distribution. In most cases the sum total of weights of successors is less than 1 which is hereby adjusted by extracting the zero weight successors for a state transition in WFA, and replacing the zero weight by allocating remaining probabilities equally as (1-sum_of_weights)/zeroelements 1440.6 Next successor in WFA state machine is chosen non-deterministically by invoking numpy.random.choice() which has the facility to return elements according to a predefined probability distribution. This is the most important prerequisite to maximize the potential of WFA synthesizer: next_random_node_np=np.random.choice(successors,1,p=wfatransitionweights_np.tolist()) 1440.7 Minutiae: An ndarray copy of WFA weights is required to circumvent the overwrite during probability distribution calculation earlier and NumPy str_ type has to be recast to Pythonn str. node=next_random_node iterations+=1 1440.8 The mixed orchestra composition is written to new file music21_piano_with_percussions.mp3 and Score tracks are printed to testlogs/AudioToBitMatrix.log.MusicSynthesizer.Music21Fluidsynth.28July2023 1440.9 This update resolves JIRA issue https://krishnairesearch.atlassian.net/browse/NEUR-33
Srinivasan Kannan
More... over 3 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 3 years ago
1439. (JIRA,THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - Concert synthesis by Percussion instrument overlay on Piano music - through Music21 MIDI,MuseScore,FluidSynth,Pydub support - 26 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434,1438 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks,Neo-Riemannian Theory of Music
Srinivasan Kannan
More... over 3 years ago
---------------------------------------------------------------------------------------------------------------------------------------------- 1439. (JIRA,THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - Concert synthesis by Percussion instrument overlay on Piano music - through Music21 MIDI,MuseScore,FluidSynth,Pydub support - 26 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434,1438 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks,Neo-Riemannian Theory of Music ---------------------------------------------------------------------------------------------------------------------------------------------- 1439.1 python-src/music_pattern_mining/AudioToBitMatrix.py has been updated to import MIT music21, hurst, pydub for composing a percussion accompaniment and mixing piano and percussion tracks to get a minimal concert audio synthesized by weighted automata which is equivalent of a 2-recurrent neural network. 1439.2 Hurst exponent of synthesized note frequencies is computed by Python Hurst which is a measure of how random a music is - as music notes oscillate between fixed set of octave notes, Hurst exponent of music is very low implying mean reversion. 1439.3 New function percussion_synthesis(virtual_piano_notes=None,percussion="TenorDrum",tempo=2000) has been defined to compose a percussion notes sequence from piano notes by MIT music21 library which offers rich set of music analysis features including neo-riemannian theory that embeds music on topological space (percussion instruments limited to 3 types of drums - SteelDrum,TenorDrum,BassDrum), set the tempo by music21 Metronome and obtain a percussion sequence of whole notes which are written to a MIDI file ./music21_percussion_for_piano.midi. This midi file can be viewed in musescore path for which is set in music21 environment. 1439.4 MIDI file synthesized for percussion earlier is translated to a WAV file music21_percussion_for_piano.wav by FluidSynth commandline invoked in os.system() and played by playsound. Percussion synthesis requires installation of musescore, music21, pydub and FluidSynth. 1439.5 Two threads are instantiated which respectively spawn executions of functions (1) music_synthesis() for piano track and (2) percussion_synthesis() for percussion track. These two tracks are overlay-ed by PyDub AudioSegment overlay to get a mixed track which is played as a concert. 1439.6 Percussion synthesis at present supports only western classical octave due to music21 limitation. This update addresses JIRA issue https://krishnairesearch.atlassian.net/browse/NEUR-32 1439.7 Following are newly added DB,synthesized midi and wav (concert - piano with percussion) music21 files along with logs: Databases.db music21_percussion_for_piano.midi music21_percussion_for_piano.wav testlogs/AudioToBitMatrix.log.MusicSynthesizer.Music21Fluidsynth.26July2023
Srinivasan Kannan
More... over 3 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 3 years ago
+1438. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - Hurst exponent of stockquote timeseries - 16 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1367,1375,1376,1436,1437 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA), Gaussian Ensemble-Gaussian Mixture Models in Weather precipitation timeseries^M
Srinivasan Kannan
More... over 3 years ago
------------------------------------------------------------------------------------------------------------------------------- 1438. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - Hurst exponent of stockquote timeseries - 16 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1367,1375,1376,1436,1437 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA), Gaussian Ensemble-Gaussian Mixture Models in Weather precipitation timeseries ------------------------------------------------------------------------------------------------------------------------------- 1438.1 python-src/Streaming_TimeSeriesData.py has been updated to import Hurst exponent python package hurst and compute_Hc() from hurst is invoked on a stockquote ticker timeseries. 1438.2 Following are the Hurst exponents from testlogs/Streaming_StockMarketData.Prophet.log.16July2023 for an example execution of python3.7 Streaming_StockMarketData.py 2>&1 > testlogs/Streaming_StockMarketData.Prophet.log.16July2023 to analyze 4 stock quotes: ------------- Hurst exponent of the stockquote timeseries ------------ Hurst exponent H: 0.6631256237890952 Hurst constant c: 0.9515361975362864 ------------- Hurst exponent of the stockquote timeseries ------------ Hurst exponent H: 0.5068001010178471 Hurst constant c: 1.3492060250154627 ------------- Hurst exponent of the stockquote timeseries ------------ Hurst exponent H: 0.5996911071760886 Hurst constant c: 0.8961089419792517 ------------- Hurst exponent of the stockquote timeseries ------------ Hurst exponent H: 0.6609766057038647 Hurst constant c: 1.0363449457436713 1438.3 Hurst exponent (*) in the range 0-0.5 implies the timeseries is mean reverting in longterm (or high value is most likely followed by low value reverting to mean in longterm) (*) in the range 0.5-1 implies the time series is in momentum and trending (or high value is most likely followed by high value) and (*) equal to 0.5 implies the timeseries is a brownian motion random walk or typical unpredictable stock defined by Efficient Market Hypothesis. Values of H > 0.5 indicate a less volatile market. 1438.4 Hurst exponent H is related to Fractal dimension D as D = 2 - H and is a measure how turbulent or difficult to predict a timeseries is.Hurst exponent encodes the long term memory of a timeseries while Fractal dimension encodes its short term memory. 1438.5 Mean reverting stockquote timeseries especially have a parallel in precipitation timeseries of weather forecast modelled as Gaussian Ensemble in Section 1367 which fit the most probable precipitation timeseries from a Gaussian Mixture Model ensemble (set of timeseries daily rainfall partitions of same mean seasonal rainfall) having NBody gravity L2 norm peaks as modes.
Srinivasan Kannan
More... over 3 years ago
Section 1431 updated
Srinivasan Kannan
More... over 3 years ago
+1437. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - Longest Repeated Substring in binary fluctuations as changepoints - 13 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1375,1376,1436 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA)
Srinivasan Kannan
More... over 3 years ago
------------------------------------------------------------------------------------------------------------------------------- 1437. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - Longest Repeated Substring in binary fluctuations as changepoints - 13 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1375,1376,1436 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA) ------------------------------------------------------------------------------------------------------------------------------- 1437.1 python-src/Streaming_StockMarketData.py has been updated to compare and contrast stock quote timeseries data for 4 tickers - AMZN,MSFT,GOOG,TSLA. 1437.2 python-src/Streaming_TimeSeriesData.py has been updated to import SuffixArray from StringSearch_LongestRepeatedSubstring.py and instantiate SuffixArray object for binary encoded fluctuations of a timeseries and find the longest repeated substring of 1s (up) and 0s (down) within it. SuffixArray of StringSearch_LongestRepeatedSubstring.py reads the binary fluctuations from python-src/StringSearch_Pattern.txt written by StringSearch_BinaryEncodedTimeSeries.py 1437.3 python-src/StringSearch_LongestRepeatedSubstring.py has been upgraded to Python 3.x by 2to3 and autopep8 and updated to import python regular expressions (re). All occurrences of longest repeated fluctuation substring and their average periodicity are computed by re.finditer() (within longest_repeated_substring() method of SuffixArray class) which are the chronological locations of changepoints in binary fluctuations (or timepoints where stock quotes for a ticker had undergone exactly similar ebb and tide). 1437.4 Python 2.7 version of python-src/StringSearch_LongestRepeatedSubstring.py has been backedup in python-src/StringSearch_LongestRepeatedSubstring.py.2.7 1437.5 Detailed prophet computation and binary changepoints from SuffixArray (Longest Repeated Substring) are at python-src/testlogs/Streaming_StockMarketData.Prophet.log.13July2023 which could be inputs to Prophet changepoints.
Srinivasan Kannan
More... over 3 years ago
Section 1431 updated
Srinivasan Kannan
More... over 3 years ago
Updated Section 1436 and Requirements.txt
Srinivasan Kannan
More... over 3 years ago
+1436. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - 11 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1375,1376 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA)^M
Srinivasan Kannan
More... over 3 years ago
------------------------------------------------------------------------------------------------------------------------------- 1436. (THEORY and FEATURE) Facebook Prophet model of timeseries data - Stock quotes usecase (extensible to any timeseries in Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations, Economics-Financial Markets) - 11 July 2023 - related to 690,691,692,1257,1315,1316,1327,1348,1375,1376 and all sections on Music Information Retrieval(MIR), Climate analytics and Weather Forecast, Computational Economics-Quantitative Finance and TimeSeries Analysis, Graphical Event Models (GEM), Causal Event Models (CEM), Chain Event Graphs(CEG), Chaos and Fractals, EventNet causality, Multifractal Detrended Fluctuation Analysis (MFDFA) ------------------------------------------------------------------------------------------------------------------------------- 1436.1 python-src/Streaming_StockMarketData.py has been updated to import and invoke newly implemented Facebook Prophet model of stockquote timeseries within Streaming_TimeSeriesData.py 1436.2 python-src/Streaming_TimeSeriesData.py has been updated to import Pandas, Facebook Prophet, Yahoo Finance Python, NumPy, Pathlib and refactored function binary_encoded_fluctuations() from StringSearch_BinaryEncodedTimeSeries.py 1436.3 New function stockquote_Prophet_timeseries_forecast_model(ticker,period='2y',interval='1wk',algorithm="linear") in Streaming_TimeSeriesData.py implements the Facebook Prophet model (defaults to algorithm "linear") for Stock quote ticker timeseries datasets obtained from yfinance. The dataset is preprocessed by to_csv() and read_csv() to adhere to Prophet() functions. The dataset must have ds and y fields in Pandas DataFrame signifying dates and values and must be stripped off timezones. 1436.4 Functions fit(),make_future_dataframe() and predict() respectively (*) fit the timeseries to a model (*) create a future dataframe placeholder and (*) makes predictions. 1436.5 Facebook Prophet timeseries model is summarized as: "...... y(t) = g(t) + s(t) + h(t) + e . (1) Here g(t) is the trend function which models non-periodic changes in the value of the time series, s(t) represents periodic changes (e.g., weekly and yearly seasonality), and h(t) represents the effects of holidays which occur on potentially irregular schedules over one or more days. The error term e represents any idiosyncratic changes which are not accommodated by the model; later we will make the parametric assumption that e is normally distributed ....." - quoted from https://peerj.com/preprints/3190.pdf ......" 1436.6 Prophet introduces the changepoints paradigm which is conceptually similar to Binary Encoding of Fluctuations in a timeseries implemented in python-src/StringSearch_BinaryEncodedTimeSeries.py through which a timeseries data is flattened to a binary string of 0s (trend is downwards) and 1s (trend is upwards), a kind of semi de-trending of a real valued timeseries. Detection of changepoints in binary encoding of a timeseries is easier - for example "11111110" substring within binary flattening of stockquote timeseries implies consecutive bullish sessions followed by a dampening event in financial markets. Trend reversals usually are associated with monetary policy decisions, blackswan events, oil prices, wars, earning forecasts, global cues among others which could be vertices of a Graphical Event Model(GEM)-Causal Event Model(CEM) or Chain Event Graph(CEG). Thus change points could also be predicted from GEM-CEM-CEGs. Binary encoding of fluctuations also helps in finding fractal self-similarity (Chaos) in non-stationary and less predictable timeseries - Stock market behaves as a Multifractal (Fractal Market Hypothesis). 1436.7 python-src/StringSearch_BinaryEncodedTimeSeries.py has been refactored and new function binary_encoded_fluctuations(ticker, period='2y', interval='1wk') has been defined encapsulating the binary-encoded changepoints of real valued stock quote ticker from yfinance. 1436.8 This code works only with Prophet 1.1.1 and holidays 0.18 as there is a regression in recent Prophet version 1.1.4 causing optimization to fail - https://github.com/facebook/prophet/issues/2354 1436.9 Following is the example binary flattened encoding (from python-src/StringSearch_Pattern.txt) of AMZN stockquote ticker fluctuations: 101000011100001011110010100001100101110000001011011011111100010011010010000001110110001011101101111111110 showing some prominent binary fluctuation changepoints (downward trend after a monochromatic streak of 4 or more upward trends) as follows (hyphenated) : 1010000111000010-11110-0101000011001011100000010110110-1111110-001001101001000000111011000101110110-1111111110- 1436.10 Stockquote data from yfinance is temporarily persisted to CSV file Streaming_TimeSeriesData.csv for Prophet preprocessing. File StringSearch_BinaryEncodedTimeSeries.py.2.7 has been backedup from 2to3 and autopep8 upgrade of StringSearch_BinaryEncodedTimeSeries.py. Logs for Prophet model forecast for AMZN ticker are captured in testlogs/Streaming_StockMarketData.Prophet.log.11July2023 1436.11 Earlier binary flucuation model retrieves all locations of changepoints when a monochromatic streak of 0s or 1s (of length >= n) reverse to 1 or 0 which could be inputs to Prophet changepoints in R - https://facebook.github.io/prophet/docs/trend_changepoints.html - R code example: m <- prophet(df, changepoints = c('2014-01-01')). Length of monochromatic streaks of 0s and 1s determines the magnitude of the trend reversal (long streak of 1s followed by long streak of 0s shows a huge market bubble followed by crash while short streak reversals are less vulnerable). Earlier Prophet model of stockquotes could be suitably adapted to any timeseries - Music, Medical Imageing-ECG, Meteorology-Precipitation, Seismology-Tremors, NBody Gravitational Accelerations.
Srinivasan Kannan
More... over 3 years ago
Section 1431 updated
Srinivasan Kannan
More... over 3 years ago
Section 1435 updated
Srinivasan Kannan
More... over 3 years ago
Section 1431 updated
Srinivasan Kannan
More... over 3 years ago
Section 1435 updated
Srinivasan Kannan
More... over 3 years ago
+1435. (THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - 7 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks^M
Srinivasan Kannan
More... over 3 years ago
---------------------------------------------------------------------------------------------------------------------------------------------- 1435. (THEORY and FEATURE) Music Information Retrieval - AI Music Synthesis by Random walks on Music Weighted Automata Directed Graph - 7 July 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433,1434 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals, Recurrent Neural Networks ---------------------------------------------------------------------------------------------------------------------------------------------- 1435.1 python-src/music_pattern_mining/AudioToBitMatrix.py has been updated to import networkx shortest_path() function which at present is limited only to printing all shortest paths. 1435.2 Function def notes_to_audio(automaton=False, function=None, deterministic=True, samplerate=44100, fractal=True, periodicity=500, weightedautomatadotfile=None,state2notedict=None,genre="WesternClassical") has been updated to include 3 new arguments namely Music Weighted Finite Automaton (WFA) graph DOT file, a dictionary defined by user which maps states of the Music WFA to 12-octave Western classical or 16-octave Carnatic piano notes scale and genre of the music to be synthesized. 1435.3 New clause has been added to read the music WFA dot file by pydot,instantiate a DiGraph() from WFA dot file and do random walk on the directed graph of WFA by lookup of state2notedict defined by user which churns out a semi-random notes sequence abiding by the automata transitions obtained by successors() function. This random walk on music WFA synthesizes new variants of a base music by randomly choosing one of the successors (which makes it an NFA) in each transition and appending a note to the synthesized notes sequence by state2notesdict lookup(s) for that state transition. The WFA music dot file could either be from machine learnt scikit-splearn Hankel matrix WFA learning algorithm or a user supplied one. 1435.4 Notes synthesized by earlier algorithm is passed on to music_synthesis() to compose a piano music waveform. This complements already implemented NotesStateMachine.txt automaton synthesis which is a user supplied finite automaton and not a machine learnt WFA. 1435.5 Fundamental objective of this code change is to complete the two way translation: 1) Music waveform to WFA by scikit-splearn and 2) WFA to Music waveform by random walks on learnt WFA directed graph. Weighted Automata have been shown to be theoretically equivalent to Second Order Recurrent Neural Networks(RNN) and RNNs are known to be Turing-complete (they can compute any mathematical function). 1435.6 2 DOT files for WFA learnt from 2 music waveforms have been instantiated as DiGraphs and new music have been synthesized from the 2 music WFAs. Music synthesized from WFA has restricted randomness and can be termed "creative" as the language generated by WFA is a subset of all possible random sequences and depends on base music waveform from which WFA was learnt - in other words WFA random walks synthesize new elaborate variants of base music or expand base music. For example following two synthesized waveforms expand learnt WFA of testlogs/054-SBC-Aanandhamridhakarshini.mp4 (Carnatic instrumental) and python-src/music_pattern_mining/testlogs/Bach_Flute_Sonata_EFlat.mp4 (Western Classical) respectively: python-src/music_pattern_mining/virtual_piano_music.Carnatic.wav python-src/music_pattern_mining/virtual_piano_music.WesternClassical.wav (especially impressive AI synthesis) 1435.7 Notes synthesized by random walks on 2 WFAs are written to python-src/music_pattern_mining/testlogs/AudioToBitMatrix.log.MusicSynthesizer.WFA.7July2023
Srinivasan Kannan
More... over 3 years ago
Section 1434 - date updated
Srinivasan Kannan
More... over 3 years ago
+1434. (THEORY and FEATURE) Music Information Retrieval - Music Weighted Automata Graph Analytics - 3 June 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals
Srinivasan Kannan
More... over 3 years ago
------------------------------------------------------------------------------------------------------------------------------ 1434. (THEORY and FEATURE) Music Information Retrieval - Music Weighted Automata Graph Analytics - 3 June 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432,1433 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity, Merit of Large Scale Visuals ------------------------------------------------------------------------------------------------------------------------------ 1434.1 python-src/music_pattern_mining/AudioToBitMatrix.py has been updated to import GraphMining GSpan implementation and NetworkX functions. 1434.2 Function music_weighted_automaton_from_waveform() has been updated to invoke music_weighted_automaton() with name of the music waveform and return the weighted automaton dotfile name. 1434.3 New function def music_weighted_automata_analytics(weightedautomata) has been defined to iterate through learnt music automaton DOT file dataset and invoke GSpan frequent subgraph mining algorithm on them for retrieving common subgraphs of music automata. 1434.4 Three music MP4 files (2 Bach Western Classical and 1 Veena Carnatic instrumental by S Balachander) are parsed to learn 3 weighted automata DOT files and music_weighted_automata_analytics() is invoked on the 3 DOT files dotfile1=music_weighted_automaton_from_waveform("testlogs/JSBach_Musicological_Offering.mp4") dotfile2=music_weighted_automaton_from_waveform("testlogs/Bach_Flute_Sonata_EFlat.mp4") dotfile3=music_weighted_automaton_from_waveform("testlogs/054-SBC-Aanandhamridhakarshini.mp4") music_weighted_automata_analytics([dotfile1,dotfile2,dotfile3]) 1434.5 python-src/music_pattern_mining/MusicWeightedAutomaton.py has been updated to change music_weighted_automaton(name=None,sample=None,splarray=None,rows=10,columns=10,thresh=0.0) to include name of the music clip and threshold (set to 0.0 by default) as configurable function arguments. The DOT and GV files are now prefixed by the name of the music clip as unique identifies. DOT strings from 3 Music Automata are written to following DOT files: python-src/music_pattern_mining/testlogs/054-SBC-Aanandhamridhakarshini.mp4_MusicWeightedAutomaton.dot python-src/music_pattern_mining/testlogs/Bach_Flute_Sonata_EFlat.mp4_MusicWeightedAutomaton.dot python-src/music_pattern_mining/testlogs/JSBach_Musicological_Offering.mp4_MusicWeightedAutomaton.dot 1434.6 Mined Frequent subgraphs of the three music waveforms can be found in python-src/music_pattern_mining/testlogs/AudioToBitMatrix.log.3July2023. It has to be noted that threshold value has to be suitably chosen to filter unnecessary transitions. In above 3 examples,weighted automata are structurally similar implying common underlying theme of any music but the weights of some of the transitions differ as could be gleaned from 3 visualizations (most transitions are of 0 weight and have to be discarded). GSpan Graph mining implementation in NeuronRain at present does not give importance to weights of the edges. 1434.7 Music datasets having high number of frequent subgraphs amongst their weighted automata representation could belong to similar music genre cluster and thus could be recommended as a playlist. The fact that Weighted Automata are already used for image compression, makes it easier to represent both audio-music and visuals (AV) as single combined weighted automata entity.
Srinivasan Kannan
More... over 3 years ago
Section 1433 updated
Srinivasan Kannan
More... over 3 years ago
Section 1433 updated
Srinivasan Kannan
More... over 3 years ago
1433. (THEORY and FEATURE) Music Synthesizer - Music Weighted Automaton Learning from music clips (user defined or AI synthesized) - 28 June 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity
Srinivasan Kannan
More... over 3 years ago
------------------------------------------------------------------------------------------------------------------------------ 1433. (THEORY and FEATURE) Music Synthesizer - Music Weighted Automaton Learning from music clips (user defined or AI synthesized) - 28 June 2023 - related to 3,110,134,298,1223,1224,1257,1258,1261,1305,1306,1307,1308,1309,1310,1311,1315,1353,1432 and all sections on Music search-Music discovery, Intrinsic Merit of Music, MIR(Music Information Retrieval), Music Similarity and Clustering, AI Music Synthesis from summation of damped sinusoids, Additive Synthesis, fMRI connectomes - OpenNeuro - Psychology of music, Weighted automata, Chaos, Deep Learning, Dynamic Time Warping-Music Timeseries analysis, Music Evoked Autobiographical Memories, Waveshaping, ThoughtNet, Majority Voting (Enjoyable music) Vs Intrinsic (Quality music) - Condorcet Jury Theorem variants,Computational Learning Theory, Boolean Function Composition, Longest Common Subsequence, Suffix Trees - Longest Repeated Substring, MultiFractal Detrended Fluctuation Analysis of Non-stationary timeseries, Ramsey Theory-Van Der Waerden Theorem for coloring of sequences, Pumping Lemma, Normalized Compression Distance-Kolmogorov complexity ------------------------------------------------------------------------------------------------------------------------------ 1433.1 python-src/music_pattern_mining/AudioToBitMatrix.py has been updated to import MusicWeightedAutomaton member functions and scikit-splearn Weighted automata SplearnArray. 1433.2 New function music_weighted_automaton_from_waveform(music_clip,duration=10) has been defined to integrate Music Weighted Automaton function music_weighted_automaton() into music analytics and synthesis (a standalone functionality so far) which enables learning automata from any music clip (user defined or AI synthesized). 1433.3 Scikit-Splearn Weighted Automaton learnt from NeuronRain synthesized WesternClassical piano notes sequence virtual_piano_music.WesternClassical.wav is written to MusicWeightedAutomaton.gv.pdf (logs: testlogs/AudioToBitMatrix.log.28June2023) 1433.4 AI synthesis or Formal language synthesis of music from learnt music weighted automata, for example, could be through (*) Creating new automata by closure (union,intersection,complementation,Kleene-star,...) of already learnt music Weighted Automata corpus (*) Use of Pumping Lemma on learnt Weighted Automata which churn out arbitrary length music notes sequences based on learnt music. 1433.5 Music similarity or Music clustering in terms of Music Weighted Automata thus reduces to their equivalence and similarity of graph underneath the automata - music could clustered by their weighted automata which is a computable regular language version of Kolmogorov complexity and compresses the dataset a la Normalized Compression Distance.
Srinivasan Kannan
More... over 3 years ago
Updated Requirements.txt
Srinivasan Kannan
More... over 3 years ago