Music105Final:Danny

From Bregman Wiki

[edit] Daniel Shapira's Final Project: FreeMapping

An blank slate, learning hardware interface that makes as few assumptions as possible about how an individual user might want to interact with the device.

A human plays (manipulates) a hardware interface that "learns" to anticipate that human's preferred "phrases" or "licks". Starts with a blank slate and slowly builds parameter mapping weights through repeated manipulations.

[MindMap of the Project]

• Overview
  • An blank slate, learning hardware interface that makes
    as few assumptions as possible about how an individual
    user might want to interact with the device.

    • Hardware continuously and seamlessly will conform its input mappings
      to what it thinks is the user's most desirable configuration <p>Alternatively, it can be a type of puzzle game where the user has to figure
      out how to control the system from a randomly generated set of initial weights </li> </ul></li> </ul> </li>

    • Implementation
      • Hardware
        • Interface as "free" as possible from pre-designed movements.  Possibilites:
          • Basic dials, sliders, buttons in "flexible" configuration
          • P5 glove like interface to free manual control
          • More to come
        • Possibly integrate 1-10 "like" feedback
      • [Software (Max/MSP and C++/Java)]
        • Supervised Neural Net (backprop) implemented on an external computer
        • Music generation algorithms implemented in Max/MSP
    • Software details
      • [Supervised Neural Net (backprop) implemented on an external computer]
        • Some possible "Like" functions
          • + If a user uses a ui element with a high granularity
          • + If a user repeated makes small backsteps over a point as if to hone in
          • + repeated actions over a small time period
          • - quick fiddle without repeat
        • Multiple training sessions???
        • Not training the y's, instead training the way x's are interacting
        • [working with Richard Granger on machine learning algorithms]
      • Music generation algorithms
    • Steps:
      • 1) Basic software implementation using software interface or P5 glove as hardware input
      • 2) Final hardware conception and design phase
      • 3) Hardware construction and software refinement

      
      

      
      

      ---