<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Teresa Marrin Nakra on Medium]]></title>
        <description><![CDATA[Stories by Teresa Marrin Nakra on Medium]]></description>
        <link>https://medium.com/@nakra?source=rss-2378cef66fef------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*w36_fN3wu4njdeD9MRSYzA.jpeg</url>
            <title>Stories by Teresa Marrin Nakra on Medium</title>
            <link>https://medium.com/@nakra?source=rss-2378cef66fef------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 06:29:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nakra/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Algorithmic Composition]]></title>
            <link>https://medium.com/@nakra/algorithmic-composition-601ed42a73d8?source=rss-2378cef66fef------2</link>
            <guid isPermaLink="false">https://medium.com/p/601ed42a73d8</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[algorithmic-composition]]></category>
            <category><![CDATA[information-retrieval]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Teresa Marrin Nakra]]></dc:creator>
            <pubDate>Mon, 06 Feb 2017 04:13:41 GMT</pubDate>
            <atom:updated>2017-02-06T04:16:58.832Z</atom:updated>
            <content:encoded><![CDATA[<p>Post 2: What is <em>algorithmic composition</em>?</p><p>In my last <a href="https://medium.com/@nakra/what-is-interactive-music-17bc5ff50915#.v8t29byh0">post on January 23</a>, I embraced the goal to define and describe interactive music systems. Now, I will take a deeper dive and look at an advanced application of interactive music systems — algorithmic composition — whereby composers employ programs to help them either better express themselves creatively or model past masters.</p><p>In order to understand algorithmic composition, we must first define what <em>algorithms</em> are. (Most of us rely on algorithms throughout our daily lives without realizing it!) You could define an algorithm as a self-contained, step-by-step set of operations or instructions to be performed by a digital device — essentially, a recipe for how to process data. The design of algorithms is a focal point for computer programmers, specifically in ensuring that their routines run as quickly and efficiently and reliably as possible. And algorithms generally require <em>parameters</em>, or special variables or pieces of data that are provided as inputs to subroutines.</p><p>But now that we know what an algorithm is, how might we apply them to music? What operations and parameters might be needed to skillfully play a flute or violin? And what algorithmic expressions could produce masterful counterpoint rivaling that of Bach?</p><p>Historical examples of musical algorithms include:</p><p>1. <strong>species counterpoint</strong>, first codified by Johann Fux in <em>Gradus ad Parnassum </em>(1725), a step-by-step method to create melodies that conform to the rules of consonance and horizontal interrelationships with fixed melodies)</p><p>2. Mozart’s <a href="http://imslp.org/wiki/Musikalisches_W%C3%BCrfelspiel,_K.516f_(Mozart,_Wolfgang_Amadeus)"><strong>Dice Game</strong></a> (<a href="https://cycling74.com/project/mozarts-dice-game/#.WI68sPNQDZE">see this modern software simulation in Max</a>)</p><p>3. 12-tone serial composition techniques</p><p>4. John Cage’s <em>I Ching</em> and other chance operation methods for composition</p><p>5. composer David Cope’s project Experiments in Musical Intelligence (EMI), (<a href="https://www.youtube.com/watch?v=yFImmDsNGdE">short documentary</a>) (<a href="https://www.youtube.com/watch?v=PczDLl92vlc">passable Bach-style chorale by EMI</a>) (<a href="http://www.radiolab.org/story/91515-musical-dna/">RadioLab story</a>)</p><p>As for a recent example of musical algorithms, meet <a href="http://www.wqxr.org/story/meet-deepbach-program-building-bach-chorales/"><em>DeepBach</em></a>, a project by Gaetan Hadjeres and François Pachet of Sony Computer science laboratories in Paris, the latest in a series of AI-based algorithmic music systems that have attempted to write music (specifically, 4-part chorales) in the style of J.S.Bach. (The scholarly preprint describing <em>DeepBach</em>’s implementation can be found <a href="https://arxiv.org/abs/1612.01010">here</a>.) <em>DeepBach</em> makes use of special algorithms called <em>neural networks</em> that model the way the human brain solves problems. It also employs <em>deep learning</em> methods (a branch of <em>machine learning</em>), which are algorithms that attempt to model and learn high-level features in datasets. Both neural networks and deep learning algorithms allow devices to learn patterns in a manner that emulates the ideas of Artificial Intelligence.</p><p>The primary contribution of Hadjeres and Pachet seems that they combine the algorithmic composition concept with machine learning and information retrieval techniques. The decision to transpose the chorales to multiple keys is an assumption that the authors make at the outset. Their reasoning is sound from an engineering perspective — to provide as large as possible a dataset with which to train their algorithms. (The authors are claiming an “agnostic” algorithmic approach, which means that they are not pre-coding assumptions about chord progressions but rather enabling their system to determine the most statistically likely harmonic, rhythmic, and contrapuntal patterns, given some starting place such as a melody.) However, the decision to transpose may not be musically sound. Although technically the keyboard was fully transposable by Bach’s time (that was the point of the Well-Tempered Clavier, after all), in actual practice certain keys were reserved for certain moods, and the vast majority of chorales are in simple keys like C, G, D, A, occasionally B-flat, F, and E or E-flat. And when choral groups sing these compositions without accompaniment, they may indeed revert to overtone or meantone tuning.</p><p>So, <em>DeepBach</em> may contain some built-in assumptions that are not perfectly aligned with what actual Bach might have done (WWBD?). But this is consistent with the claim of Music Information Retrieval and machine learning methods that by going with a pure, “unsupervised” modeling approach, they are avoiding some of the biases that come along with human-generated algorithms. It is the belief of some of the researchers in this area that the computers are uncovering hidden structural elements in our music that we ourselves have not yet been able to understand. (My own view is somewhere in the middle — yes, these techniques are powerful, but without the subtle nuances that human musicologists and theorists and composers understand, the results might be ultimately musically uninteresting and without actual cultural value.)</p><p>[many thanks to Don Hart for recently bringing DeepBach to my attention!]</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=601ed42a73d8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What is Interactive Music?]]></title>
            <link>https://medium.com/@nakra/what-is-interactive-music-17bc5ff50915?source=rss-2378cef66fef------2</link>
            <guid isPermaLink="false">https://medium.com/p/17bc5ff50915</guid>
            <category><![CDATA[music]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[new-music]]></category>
            <category><![CDATA[tech]]></category>
            <dc:creator><![CDATA[Teresa Marrin Nakra]]></dc:creator>
            <pubDate>Mon, 23 Jan 2017 16:17:38 GMT</pubDate>
            <atom:updated>2020-03-29T21:08:25.954Z</atom:updated>
            <content:encoded><![CDATA[<p>For my first contribution to Medium, I thought I should address a topic that is relevant to a course I teach at The College of New Jersey: <a href="http://www.tcnj.edu/~nakra/courses.html"><em>Interactive Music Programming</em></a>. In this class, we learn how to write computer code that enables us to interact meaningfully with music. Every semester, my students present their interactive music systems for the public. These performances feature new compositions, improvisations, arrangements of existing works, and delivery systems for music learning.</p><p>But before getting into how we make all of this possible, I need to first establish what interactive music is! After all, most music is highly interactive already. So what does it mean to create <em>interactive</em> music? According to Robert Rowe (<a href="https://wp.nyu.edu/robert_rowe/text/interactive-music-systems-1993/chapter-1-interactive-music-systems/">Interactive Music Systems: Machine Listening and Composing. MIT Press, 1992, page 1</a>): “Interactive computer music systems are those whose behavior changes in response to a musical input.” According to this definition, merely pressing play on your favorite playlist is not interactive. So what are the minimum requirements for interactivity? What counts as <em>responding</em> and <em>changing</em> based on input?</p><p>For music to respond and change flexibly to input, we need to use the technique of <em>musical parameterization</em>, or the creation of abstract, numerical structures from procedural descriptions and algorithms. The better the algorithms one can write, the more intelligent, autonomous, and responsive their systems will become. Ideally, such systems don’t just spit out streams of robotic notes, but also can improvise, compose, and perform with sensitivity, fluency, and emotion. And they can function at the level of description and cognition required by the human needs of the situation at hand.</p><p>Here are some examples of interactive music systems that I admire:</p><p>· Zoe Keating (cello &amp; electronics with Max), <a href="http://www.youtube.com/watch?v=yYrcXX4nWOA&amp;feature=related">Escape Artist</a></p><p>· Smule’s <a href="https://www.youtube.com/watch?v=qPqXxmC51bI">Ocarina app</a> for the iPhone</p><p>· <a href="http://www.smartmusic.com/">Smartmusic</a> (by Roger Dannenberg)</p><p>· <a href="https://en.wikipedia.org/wiki/Triadex_Muse">MUSE synthesizer</a> (invented by Marvin Minsky and Ed Fredkin — a device for not just generating sound, but composing and improvising by creating streams of note patterns, allowing the user to set the parametric relationships via sliders and buttons.)</p><p>As our development platform, we use Cycling ’74 Max, MakeyMakey kits, and a range of other sensors, controllers, and devices to help us discover the possibilities for making interactive, responsive music that we all can enjoy! Here’s a video of one of our <a href="https://www.youtube.com/watch?v=Oj4q_z7_xG8&amp;t=25s"><em>Music to the Max</em></a> class performances. Stay tuned…and keep an eye out for my next post on interactive <em>conducting</em> systems!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=17bc5ff50915" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>