Bach In Beta: Modeling Bach Chorales With Markov Chains

2y ago
13 Views
3 Downloads
1.35 MB
73 Pages
Last View : 1d ago
Last Download : 2m ago
Upload by : Sasha Niles
Transcription

Bach in Beta: Modeling Bach Choraleswith Markov ChainsMichaela Suzanne TracyA ThesisPresented to the Facultyof Harvard Universityin Candidacy for the Degreeof Bachelor of ArtsRecommended for Acceptanceby the Department ofApplied MathematicsAdvisers: Professor Mauricio Santillana and Professor SuzannahClarkMarch 2013

c Copyright by Michaela Suzanne Tracy, 2013.All rights reserved.

AbstractIn this thesis I present a Hidden Markov Model capable of composing musical choralesinspired by the works of J.S Bach. The transition matrix for the model is constructedby sequentially incorporating musical information from approximately seven percentof the 371 chorales composed by J.S. Bach. The quality of the chorales producedby the computational model was tested experimentally by surveying 40 musical experts. Our results show that musical experts could not identify the differences between chorale harmonies composed by J.S. Bach and synthetic harmonic progressionsproduced with our algorithm. By performing statistical hypothesis testing on the proportion of correct responses, we found that the success rate of identifying an actualBach chorale correctly, among experts, was similar to flipping a coin for a transitionmatrix trained with more than 10 chorales. This finding suggests that the computational methodology presented here is a successful way to create Bach-quality harmonicprogressions.iii

AcknowledgementsI would like to thank my advisers, Professors Santillana and Clark, for giving me theattention and guidance in order to complete this project. Also, to Simon Lunagomez,for helping us design the experiment, and all his wonderful feedback throughout theprocess. I would like to thank my parents, whose laid-back parenting complementedwith loving support motivates me to achieve all of my ridiculous goals. To Dr. KevinC. Leong, whose sheer love of learning, teaching, and music have inspired me since myfreshman year. His friendship can never be replaced, and I cannot imagine how I wouldhave survived the last years without his substitute-parental concern. I also thankNicandro, who was my first inspiration to take on this project, and our whirlwindfriendship that never ceases to surprise me. Finally, to my brother, all the respect inthe world for his perseverance and dedication to the things he loves.iv

To KL, NI, ’Olden and my family.v

ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ix1 Overview12 Musical Introduction33 Mathematical Introduction84 Model135 “Original” Method176 Improved Model317 Experiment377.1Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . .377.2Experiment Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .397.3Testing the Null Hypothesis . . . . . . . . . . . . . . . . . . . . . . .397.3.1Analytical Approach 1 . . . . . . . . . . . . . . . . . . . . . .397.3.2Analytical Approch 2 . . . . . . . . . . . . . . . . . . . . . . .417.3.3ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42vi

7.4Exploratory Data Analysis . . . . . . . . . . . . . . . . . . . . . . . .437.5Additional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . .437.6Musical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .458 Application to Education489 Conclusion539.1Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .549.2Take-Aways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .55A Code56A.1 MATLAB code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .56A.2 Mathematica code . . . . . . . . . . . . . . . . . . . . . . . . . . . .57A.3 Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59Bibliography63vii

List of Tables3.1Theoretical Weather Markov Matrix. . . . . . . . . . . . . . . . . .84.1Stationary Distribution of the Last Chords of G Major Chorale . . .145.1Transition Matrix from Minor Chorale 1 . . . . . . . . . . . . . . . .195.2Transition Matrix Based on Minor Chorale 2 . . . . . . . . . . . . . .195.3Transition Matrix Based on Major Chorale 1 . . . . . . . . . . . . . .205.4Transition Matrix for Major Chorale 2 . . . . . . . . . . . . . . . . .205.5Steady State for Minor Chorale 1 . . . . . . . . . . . . . . . . . . . .235.6Stationary Distribution of Major Chorale 1 . . . . . . . . . . . . . . .245.7Transition Matrix for Minor Chorales Combined . . . . . . . . . . . .245.8Transition Matrix for Major Chorales Combined . . . . . . . . . . . .255.9Stationary Distribution of Minor Chorale Combination . . . . . . . .265.10 Stationary Distribution Major Chorales Combo . . . . . . . . . . . .276.1Transition Matrix Major Chorale 1 . . . . . . . . . . . . . . . . . . .346.2Analytic Stationary Distribution of Major Chorale 1 . . . . . . . . . .367.1Mean Fractions of Correct or Incorrect Responses from the Experiment, Training Sets 1-8 . . . . . . . . . . . . . . . . . . . . . . . . . .7.243Continuation of Table 1 for Data from Training Sets 12-25 (StandardDeviation in Parenthesis) . . . . . . . . . . . . . . . . . . . . . . . . .viii44

List of Figures2.1Ratio between pitches seperated by the 12 semitones . . . . . . . . .42.2C major and A minor scale . . . . . . . . . . . . . . . . . . . . . . . .52.3The seven chords based on E-Flat Major . . . . . . . . . . . . . . . .52.4The seven chords based on G minor . . . . . . . . . . . . . . . . . . .62.5Example of an Open Root Position Chord . . . . . . . . . . . . . . .73.1Graph of Stationary Distribution . . . . . . . . . . . . . . . . . . . .114.1Available Diatonic Chords of G-Major in Root Position . . . . . . . .134.2Closing Chords of Brich, an. Chorale in G major . . . . . . . . . . .144.3Stationary Distribution after 196 time-steps . . . . . . . . . . . . . .155.1Stationary Distribution of Minor Chorale at 12 time-steps . . . . . .225.2Stationary Distribution of Major Chorale at 18 time-steps . . . . . .235.3Stationary Distribution of Minor Combo at 11 time-steps . . . . . . .255.4Stationary Distribution of Major Combo at 11 time-steps . . . . . . .265.5Last Four Bars of Major Chorale 2 Trial . . . . . . . . . . . . . . . .285.6Bars 6-9 of Minor Chorale Combo Trial . . . . . . . . . . . . . . . . .306.1KERN Data Format . . . . . . . . . . . . . . . . . . . . . . . . . . .336.2Improved Method Major Chorale 1 Stationary Distribution . . . . . .35ix

7.1Box Plot of the Correct Responses Fraction vs. Number of Choralesin the Training Matrix . . . . . . . . . . . . . . . . . . . . . . . . . .7.240Relative Percent Error of Transition Matrix in Relation to the M25Transition Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44Box Plot of Experience, per level of training . . . . . . . . . . . . . .45A.1 Opening of a Lesson Plan . . . . . . . . . . . . . . . . . . . . . . . .59A.2 Unit Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .60A.3 Sample Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .60A.4 Sample Worksheet . . . . . . . . . . . . . . . . . . . . . . . . . . . .61A.5 Sample Homework Question . . . . . . . . . . . . . . . . . . . . . . .61A.6 Sample Test Question . . . . . . . . . . . . . . . . . . . . . . . . . . .627.3x

Chapter 1OverviewThe relationships between music and mathematics are natural to make, and havebeen made for centuries. From the determination of the mathematical ratio betweenaudible pitches, to the numerical classification of pitches, and the regular summationof rhythmic events in musical measures, musicians use mathematics consistently inalmost every realm of study, practice, and performance. In addition to the directimpact mathematical relationship and understanding have on music, mathematics canalso be indirectly applied to music in order to gain better understanding. In the early18th century, the ideas of quantifying music in terms of numbers became a soughtout practice. Groups of sounds began to be named in terms of numbers as opposedto musical letter names, and numerical relationships were described academically.Into the 20th century, composers began to use mathematics as the spark for musicalideas, such as replicating numerical patterns in their music. These less-consistentapplications nonetheless inspire musician-mathematicians to find new ways to usemath in music.Before these secondary uses of mathematics were even formally addressed, J.S.Bach composed much of the work best known for inspiring formal music theory andthe mathematics it implores. Bach (1685-1750), who lived his entire life in Germany,1

composed music of nearly every genre in great amounts. His contributions to solorepertoire, small and large-groups, and choirs are innumerable. One of the greatestcontributions to Western music was Bach’s chorale harmonizations. While he wasdirector of music in Lutheran churches, he took the standard Lutheran hymns andreworked them for four voices, with one voice retaining the original melod. Bach usedthese four-part chorales in larger works, such as cantatas or oratorios. Today, 371 ofthese chorale harmonizations still survive.What these chorales have provided for music theorists and historians is a largedata-set of music from the Baroque era of Western music. By studying these works,we can tell how often Bach used certain musical devices in order to gain better understanding of the music. We can also study patterns, frequency, and dependency ofevents using mathematics. Applying the idea of Hidden Markov Models to the largedata set of chorales, the possibilities of non-trivial results seem endless. However, certain results are more critical to the understanding of harmony, such as the regularityof chords and how replicable chorales are via statistical and probabilistic means.2

Chapter 2Musical IntroductionThere is a great deal of vocabulary from the musical lexicon that will be used in thisproject. Hopefully, the exact musical significance of each term will not be necessaryto understand the mathematical model and its construction. Despite the fact thatthere is no need to know all aspects of music theory to understand this project, it willbe important to understand a few key concepts in order to understand the significanceof the results. First off, a pitch (or note) has a specific frequency described in hertz.There are 12 pitch classes in Western musical notation. For example, the note A,which most modern orchestras tune their instruments to, has a frequency of 440Hz.When are person plays a note exactly twice this frequency, or 880Hz, that note issaid to be ”an octave” higher. Notes an octave apart are given the same pitch name,but are distinguished by their frequency, or the name of the octave.The octave is then divided into 12 intervals, known as semitones. An octave isa musical distance between two pitces that is 12 semitones large. The pitch A440 isin the 4th octave, as where A880 is in the 5th octave. Besides the octave, there areseveral other important names to intervals. Two notes with identical frequencies arecalled unisons. The pitch that is in a 3:2 ratio to the original frequency is called the”perfect fifth.” The octave, unison and perfect fifth are considered the most funda3

mental intervals in music. The following table describes the remaining relationshipsbetween the 12 pitches.Figure 2.1: Ratio between pitches separated by the 12 semitones [11]A key is a set of pitches (when played in order called a scale) that have predetermined spacing starting from the root of the key. There are two types of keys andscales, major and minor keys. Major scales start on the root pitch and then the other6 notes are 2,4,5,7,9, and 11 semitones above the root pitch. A minor scale beginswith the root pitch and has 6 other notes 2,3,5,7, 8 and 10 semitones above the root.Since there are twelves possible notes and two types of key, there are 24 possible keysthat can be explored in this project. The important advent of the music of J.S. Bachis that it uses all 24 keys.Chords are sets of three or more specific pitches, or notes, played simultaneously.In this project, only chords of three of four notes were studied, since they are thesignificant majority of all chords in Western Baroque and Classical music. Chordsare given names based on which pitch the chord is “based on.” This fundamentalnote of the chord is called the root. Chords of three notes can be if three positions:4

Figure 2.2: C major and A minor scale, who share the same pitches different order [5]one where the root is the lowest frequency note in the chord (root position chord),another where the root is the middle frequency note (first inversion chord), and onewhere the root is the highest frequency note (second inversion chord). In this study,only a few types of chords have four notes, and they occur in four positions with thesame naming system. Because of the twelve possibly notes and seven types of chordsin inversion, there are 84 possible chords in Western harmony (to be studied in thisproject). There exists countless other types of chords, but these 84 types will bethe bulk of the analysis of this project. Despite the great range of possibility, mostpieces restrict themselves to 15-20 different chords per piece. Some chords never occurin Western music because of their perceived “dissonance” or “bad-sounding” naturecompared to the tonic of a key, or the preceding chord.Figure 2.3: The seven chords based on E-Flat Major [3]In music, we give chords Roman Numerals based on their relationship to the tonic.The tonic chord, based on the tonic and two notes above it in the major key, is labeled5

Figure 2.4: The seven chords based on G minor [3]I. The second chord is called ii, and this pattern continues on. The Roman numeralis capitalized if the chord is a major mode, and is lower-case if the chord is a minormode. In figures 2.3 and 2.4 we see the diffence in Roman Numeral naming in Majorand Minor keys. These are the significant, three-note root position chords used in thisproject. Chords can occur in a variety of spacings, as well (since there are multipleoctave-equivalent notes in music). The chords displayed in these figures are calledclosed-position chords. The definition of a closed-position chord is that no chord tones(notes belonging to a specific chord) are skipped or omitted from the starting pitchof the chord to the ending pitch. A closed-position chord may omit, for example, thesecond note of the chord and have it occur an octave higher or lower. The figure 2.5below shows an F-major chord (the IV chord of C major, since it is based on the 4thpitch of the C major scale and its pitches classify this chords as being major) in anopen position to contrast to all the closed position chords in Figures 1.3 and 1.4. Themost important Roman Numerals to remember are the I (or i in minor), as this isthe tonic chord of the key. The other most important chord is the V (almost alwayscapitalized in major and minor keys) as this chord is known as the dominant, andas its name suggests it has a special function. The dominant usually functions as amusical signal that the tonic is arriving, and has a musical feeling associated withwanting to ”go somewhere” or return home to the tonic.Sometimes, there are special classifications of chords whose pitches do not comefrom the scale or key that the piece is in at that time. These chords are known asapplied chords. For example, if a section of a chorale is in the key of C major, any6

Figure 2.5: Example of an Open Root Position Chord [6]pitches with sharps or flats are not part of that key. A common chord to see in theseC major sections may be a D-major chord, which contains an F-sharp. This chord isknown specifically as a secondary dominant, as D-major is the V or dominant of Gmajor, (and G major is the dominant of C major). Secondary dominants can stackup onto each other and are noted as V/V, with the first Roman numeral describingthe chord type, and the second Roman Numeral representing what the key or scalethe chord belongs to.Finally, the last idea integral to the overview of musical ideas in this project isthe concept of musical harmonic progressions. A chord progression is the sequenceof chords that occur in a piece or section of a work. A cadence is a musical pause orclosing the ends an idea. In this era of music, most of the important cadences end withV-I in root position or inversion. Between an opening and a cadence, there is muchmore freedom for different harmonic events. The basic framework for a progressionis tonic (I)-predominant (any type of chord, but frequently ii, IV and vi in major)dominant (V)-tonic (I). This framework will be discussed as it relates to the resultsof the modeling experiment in subsequent chapters.At this point, there is a baseline level of musical understanding in order to understand the model. The goal of this project is to catergorize the different chords thatoccur in the Bach Chorales, separating for major and minor, and normalizing for key.7

Chapter 3Mathematical IntroductionThe mathematics behind modeling music is frequently used for many other topics.The main scope of this project deals with Markov processes, or the mathematicsbehind networks of interaction. A Markov chain is a mathematical system, generallyin matrix form, that models transitions of a given phenomenon. The underlyingconcept of Markov chains is that the method shows the likelihood of moving to astate, or an event occuring, given we know our starting state, or previous event. Thequintessiantial introductory example of a Markov chain is a theoretical example aboutpredicting the weather given that the observes knows the weather of the current day.1A sample matrix of this situation may look like this:Table 3.1: Theoretical Weather Markov MatrixRain Sunny SnowRainSunnySnow0.40.30.40.40.60.30.20.10.3In this example, there are three state: raining, sunny or snowy. The way the tableis interpretted is that the current state (or weather is given by the row.) The columngives the probability that the next state is the given column name. For example, the1I was introduced to the concept of Markov Chains with a similar example in Applied Math 115.8

second row, first column number describes the probability that if we know today issunny, the probability tomorrow will be rainy is 0.3.There are many important terms when it comes to Markov Chains and their analysis. One of the key ideas is reducibility. Reducibility is the idea that certain statesmay not be possible to ever transition to given a specific starting point. Accessibilityis the concept that if you start in state ”j” that through some amount of indeterminate time-steps it is possible to arrive in state ”k” (therefore, state ”k” is accesssiblefrom state ”j”. ) [4]States are also known to be commutative if it is possible to go from state j to kand k to j with varying probabilities. A commutative class is a group of states whereall other states are accessible and commutative, and no other states in the class cancommuticate with any states outside the class. Relating back to reducibility, a Markovchain is said to be ”irreducible” if its entire state-space is one commutative class[4].After the transition matrix of the Markov chain is created, we can analyze it tolearn about its important characteristics. A stationary distribution describes thatthe long-term probability of the states occuring. The stationary distribution canbe solved in a few different ways. The first way is to set up a system of equationsbased on the matrix. In this case, there are three variables and three equations. Forsimplicity, let R rain, S sunny, and W snow. We also have to add the constraintthat the total probability is 1, R S W 1, (that is, there is a probability of 1 thatthere is some kind of weather on a given day).Equation 1: R .4R .3S .4WEquation 2: S .4R .6S .3WEquation 3: W .2R .1S .3WEquation 4: 1 R S WThese equations can solved in a variety of ways, the easist method being usinglinear algebra to solve the Ax b situation. One could also do elimination/substitution9

method for the three variables. Either way, the solution of these equations are: [Rain 0.352112676248874, Sunny 0.478873238932554, Snow 0.169014084818572].Solving a system of equations with multiple variables can be very time consuming.A quicker way is to use Markov theory to obtain the steady-state distribution. If wecontinue to apply the transition matrix to an arbitrary vector, then eventually thevector will converge to the stationary distribution. In this example, say we want todetermine the distribution given we start in each state (separately). We can use thefollowing MATLAB algorithm to continue to multiply the transition matrix until thedifference of the norms between steps is within .001.If a is the transition matrix above:w(1) 1bar(w)ylim([0,1])pausefor k 1:15t w*a;if norm(w-t) .001, break, endw t;bar(w)pauseendAfter 7 steps, the matrix converges to the exact same probabilities from the analyticsolution. The graph of transition probabilities is shown in figure 3.1.What this diagram hopes to show is that no matter what state one starts in, overtime (7 intevals of time, or days in this case) the probability of the specific next stateoccuring is the same.Beyond the graphical, iterative analysis of the steady-state distribution, this analysis can also be done by using eigenvectors. Eigenvalues are normally defined by theequation Ax λx or (A λI)(x) 0 , where λ is the eigenvalue, x is the eigenvector,A is the matrix one is trying to find the eigenvectors of and I is the identity matrix.10

Figure 3.1:Stationary Distribution of Weather Problem with the Aformentioned AlgorithmIn the rainy, sunny, snowy example, there is an interesting concept that surfacesthat is normally not discussed in basic Linear Algebra. Because Markov Chains, andthis problem, deal with row vectors multiplied by the transition matrix (usually thematrix is multiplied first) the idea of eigenvectors has to be altered. The aforementioned eigenvector equation solves for column vectors, or the right-eigenvectors of thematrix. In the case of Markov Chain, since we are multiplying in a different order,we need to solve for left-eigenvectors, or rows. The equations for left-eigenvectors arexA λx or (AT λI)(xT ) 0The analytical way to solve for eigenvalues (the λ constants in the eigenvectorequations) is to solve for the eigenvalues and solve for the null-space of the matrixcorresponding to that particular eigenvalue. This process solves for the λ’s and thensolves the second form of the equation for the x term, which is simply the eigenvector.The simplest way to solve for the eigenvalues is to set up the characteristic equationfor the matrix. The characteristic equation says that we want to find the valuesof λ that when subtracted from the original matrix, make the matrix singular, and11

give the matrix a determinant of 0. The equation for the characteristic equation isdet(A λI) 0. In the rainy, sunny, snowy example, we have a 3X3 square matrixwhere the characteristic polynomial would look like:.4.2 .4 λ det .6 λ.1 .3 0 .2.1.3 λBy using determinant rules we can determine the eigenvalues of the matrix andthen apply them to the left-eigenvalue formula. This method yields identical resultsto the other previously mentioned methodsThough the Markov model presented in this project is not often labeled this way,this process is more closely related to a Hidden Markov Model. A Hidden MarkovModel (or HMM) is a process where the method involved is an underlying stochasticprovess that can only be observed through another set of stochastic processes.[7].Thus, in our model, we are trying to figure out Bach’s personal harmonic model bythe observable events of his chorale progressions. We learn more about the choraleharmonies in order to learn about Bach’s probabilistic model of harmony. Embeddedin the model is an overarching discovery of this HMM, and though we often referto the model as a regular Markov chain, in reality we are also learning about thisintegrated hidden model.With a basic understanding of how Markov Chains (hidden or not) work andhow stationary distribution can be calculated, we can now begin to think about howMarkov Chains can be used to describe chord progressions in Bach Chorales. Thesetools, such as solving for stationary distribution using iterative code and eigenvalueswill be critical to the analysis of the method in this project.12

Chapter 4ModelIn this chapter, we describe how we constructed the mathematical model for thismusical phenomenon. Let us first look at a section of one Bach Chorale that wasused to train the original model of this project. This chorale is a major-mode choralefrom Bach’s Christmas Oratorio (Part 2) entitled Brich an, O Schönes Morgenlicht(Break Forth, O Beauteous Heavenly Light). The chorale is in G-major, and thuswould have the following seven chords available, with each root of the chord beingone of the pitches from the G-major scale.Figure 4.1: The seven chords based on G major [3]With this set of available chords, we can begin to think about applying the conceptof transition states to the chorale. Let us first look at the last section of the chorale,including the important, final cadence. In purple are the Roman Numerals for thissection.13

Figure 4.2: Closing Chords of Brich, an. Chorale in G major [2]This closing section is very straightforward, and attempts to show how systematiccadences are in Baroque harmony. The final two chords of the piece are V and I (tonicand dominant) and they are preceded by a pre-dominant ii61. We see that in thisexample, the only state that does not have a certain state to follow is the V chord,which can transition to I or I6 with equal probability. When we determine the steadystate of this situation using the left-eigenvector, and normalize so that the sum of allelements of the steady state is 1, we get a very regular fluctuation between .1 and .2probabilities.Table 4.1: Stationary Distribution of the Last Chords of G Major ChoraleChord A chord in first inversion is called X6, second inversion is called X6/4.14

However, when we try and solve for the stationary distribution using the iterativemethod, it takes exactly 196 times steps for the model to completely converge to thestationary distribution for all starting states of the chain. In the weather example inthe introduction, the 3X3 matrix converged in 7 time-steps. One of the reasons whyit takes much longer for this first Bach matrix to converge, as opposed to the weathermatrix, is the size, length, and sparseness of the matrix. Note that many states areonly attainable from one preceding state, and though the Markov chain is irreducible,it is still difficult to reach some states if you start in the state that is multiple degreesof accessibility away from the given state.Figure 4.3: Stationary Distribution after 196 time-stepsWhat this initial set up also tells us is that this amount of data will not suffice inhelping us create an idea for the Markov Chain. We need much more data in the hopes15

of creating a usable model. However, what this example shows is what we are trying tomodel. By taking the Roman Numerals, dividing up the occurances of each transition,and solving for the steady state, we are able to learn some important characteristicsof this chain. First off, the three most likely states, with equal probabilities in thesteady state analysis, are the I chord, the ii6 chord, and the V chord. Initially, itmakes sense that I and V are most likely, since they are the only two chords in thepassage that appear more than once. However, it is interesting to discuss why ii6 isalso equally likely in this situation. In the steady-state analysis, contrary to intuition,the most popular or most frequent states in the training matrix do not always lead tothe most likely state in the stationary distribution. The reason why ii6 is so probableis that is it well-connected to states such as I and V, which occur quite frequently.The model, in principle, is supposed to represent all of Baroque harmony. Clearly,a set of 8 states (chords) in this case are not enough data to quantify all of Baroqueharmony. Further analysis is needed to determine what types of patterns are commonfor Bach chorales. Also, this initial analysis highlights the important aspects ofMarkov chains in Bach chorales. The Markov chains in these chorales are irreducible,and therefore, every state is significant. Roman Numeral analysis is also useful inthis setup because it describes relationships between chords and pitch classes, but isapplicable to all keys. Since Bach’s music uses almost every key, using chord symbolsand relationships as opposed to notes is more useful. Overall, this method is definedto help solve the issues of harmony, and hopes to quantify a fully-accepted qualitativeanalytical method.16

Chapter 5“Original” MethodFrom the model setup in Part 4, we can see that this method has much to offerin terms of insight and analysis. Some limitations with our approach may arise ifthe constructed transition matrix is ill-conditioned. To overcome the ill-conditionedproperties of the matrix, we must be sure to categorize the data appropriately. Whatthis method is designed to achieve is information about the implicit Markov chains inBach’s chord progressions. By learning more about these progressions, we can thentrain a model with a set of raw data from the Bach chorales. With this data, we canrun Monte Carlo trials (random trials using

Bach used these four-part chorales in larger works, such as cantatas or oratorios. Today, 371 of these chorale harmonizations still survive. What these chorales have provided for music theorists and historians is a large data-set of music from the Baroque era

Related Documents:

J S Bach 'Komm, Jesu, komm' J C Bach 'Lieber Herr Gott, wecke uns auf' J S Bach 'Lobet den Herrn, alle Heiden' interval 20 minutes J C Bach 'Herr, nun lässest du deinen Diener' J S Bach 'Jesu, meine Freude' J C Bach 'Der Gerechte, ob er gleich zu zeitlich stirbt' J S Bach 'Singet dem Herrn ein neues Lied' Solomon .

Jacksonville, FL ALPHA DELTA Miami, FL ALPHA EPSILON Atlanta, GA ALPHA THETA New Orleans, LA ALPHA OMICRON Tampa, FL ALPHA PHI Pensacola, FL BETA BETA Montgomery, AL BETA ETA Memphis, TN BETA KAPPA West Palm Beach, FL BETA XI Orlando, FL SOUTHEAST REGION CHAPTERS BETA SIGMA Lakeland, FL BETA PSI Knoxville, TN

Bach JS Italian Concerto 1st mvt 8 Bach JS Italian Concerto 3rd mvt Presto 8 Bach JS Little Prelude in Dm BWV 940 5 Bach JS Little Prelude in No 4 in D BWV 936 6 Bach JS Overture in F BWV 820:5 Bourree 4 Bach JS Partita No 1 in B flat BWV 825 Praeludium and Giga 8 Bach JS

Alpha Kappa, SW Alpha Chi, SW Beta Zeta, SW Beta Kappa, C Beta Mu, S Beta Rho, SE Beta Upsilon, S Gamma Gamma, SE Bea Davis Class of 2015 Pam Hesser, Mu, NW Glenda Spurgeon, Beta XI, C Ivy Trozier, Alpha Upsilon, C Pat Crochet, Iota, S Marjorie Bull, Mu, NW Sherry Fulton, Psi, C Emily Armold, Theta, NE

0.9721 0.014. This value of beta is 33% lower than the antarctic beta diversity. The additive beta definition fails to rank these data sets correctly because the beta it produces is confounded with alpha. (When diversity is high, Gini-Simpson alpha and gamma both approach unity. Therefore if beta is defined as gamma minus alpha, beta must .

Bach, JS Two-part Invention No. 4 in D minor BWV 775 5 Bach, JS Two-part Invention No. 8 in F BWV 779 5 Bach, JS Two-part Invention No.14 in B flat BWV 785 6 Bach, JS arr. Keveren Air on the G String 6 Bach, JS trans. Alkan Siciliano 7 Bach, WF Aria in G minor 4 Bacharach Raindrops Keep Falling On My Head 4

Bach himself. Many of Bach's chorales are harmoniza-tions by Bach of pre-existing melodies (not by Bach) and certain melodies (by Bach or otherwise) form the basis of multiple chorales with different harmonizations. We extend this harmonization task to the completion of chorales for a wider number and type of given parts. Let

EMR 911L BACH / GOUNOD Ave Maria EMR 902L BACH, Johann S. Aria EMR 913L BACH, Johann S. Arioso EMR 2104L BACH, Johann S. Chorale Prelude "Ich ruf zu Dir" EMR 217L BACH, Johann S. Jesu, meine Freude (Reift) EMR 8474 BACH, Johann S. Lobe den Herrn (5) EMR 21