CarianiHomePage.html

Recent events Jan-March, 2008)

Presentation to session on “Flattening Federal research funding: The local angle”, Annual meeting of the Association of Health Care Journalists, Alexandria, VA, Friday, March 28, 2008, "The NIH Flat-funded: A view from the ground level. (PDF).

Scholarpedia article on "The Jeffress Model". (PDF) Article in progress, comments are welcome.

Recent events: Symposium on "Creativity: The Mind, Machines, and Mathematics", MIT Stata Center, November 30-December 2, 2006, sponsored by the Templeton Foundation. My working paper for this discussion: "On creating new infomational primitives in minds and machines", PDF.

I am still very much interested in the neural coding problem, temporal codes, neural timing nets, auditory neurocomputation, auditory scene analysis, the neural representation of pitch, and the neural basis of music perception.

I firmly believe that if we are to successfully understand how the brain works as an informational system, we need to understand the precise nature of the pulse-coded signals it uses. The neurosciences desperately need to address problems of neural coding head-on, and we need to contemplate neurocomputational alternatives to rate-codes and traditional connectionist networks. One can glimpse the power of temporal pulse codes and computations that could permit neural signals to be liberated from dedicated transmission lines, much in the same way that radio and internet have transcended telegraph networks. The high dimensionality of temporal pattern codes affords more flexible means of implementing type-based logics in neural networks. I am interested in seeing these new kinds of temporal processing networks become a reality; please contact me if you know of persons or organizations interested in funding such paradigm-shifting work.

In the Fall of 2003, Spring of 2004, and this past semester, I had the privilege of organizing and teaching courses on music perception and cognition at Tufts University and MIT. I will be teaching HST 725 Music Perception & Cognition again this spring at MIT, TR 7-9 PM, E25-101. The MIT course, HST 725, has an OpenCourseware website from 2004 that is freely accessible to all (URL below). A good deal of new material was been added for 2007 (email me if you are interested in current course content).

http://ocw.mit.edu/OcwWeb/Health-Sciences-and-Technology/HST-725Spring2004/CourseHome/index.htm

As a personal favor to Mary McKinnon, who taught my son piano until serious illness intervened, I have scanned The Capture of Imagination by her mentor, E. Robert Schmitz. The book, published in 1935 by Carl Fischer and long out of print, is a rational theory of piano performance based on the mechanics and kinematics of human and instrument. Mary is one of the last surviving students of Schmitz, and she is writing a commentary on his pedagogical method. Our understanding is that the publisher has no current interest in reprinting the work, and does not mind if an electronic copy is made freely available to the world. PDF and .doc formats can be downloaded: Schmitz (1935, pdf, 32.4 Mb).

End of update 6.03.07

Keywords: auditory neuroscience, theoretical neuroscience, theoretical biology, biological cybernetics, epistemology, philosophy

Which aspects of neural response convey functionally-relevant information?
Patterns of spikes vs. patterns of channel-activations

            General theory and examples from different modalities
            The many consequences of phase-locking
            Extrinsic, stimulus-drive time structure vs. intrinsic, stimulus-evoked time structure

Temporal discharge patterns in the auditory nerve. When a periodic sound is presented to the ear, it impresses its own temporal structure on the discharges of auditory nerve fibers. As a consequence of this stimulus-driven time structure, there exist correlations between timings of spikes and the stimulus waveform ("phase-locking"). In this case the stimulus is a single-formant vowel, with a fundamental frequency of 80 Hz and most spectral energy in harmonics near 640 Hz (formant frequency), presented 100 times at 60 dB SPL. The stimulus produces a strong "voice pitch" at its fundamental, and this periodicity can be seen in the discharge patterns of auditory nerve fibers across the auditory nerve array. When time durations between both successive and nonsuccessive spikes are tabulated in an interspike interval histograms, intervals related to the fundamental period (pitch period) and to spectral shape (tone quality, timbre) are seen. For harmonic stimuli, the most frequent intervals in the auditory nerve are those associated with the fundamental, such that the pitch that is heard invariably corresponds to the most frequent interspike interval. Exceptions to this rule involve stimuli that produce buzzy, rate-pitches an octave above the fundamental as well as some subtle pitch shifts that can occur with large changes in level.

The figure on the right shows two possible neural representations of the stimulus. The two plots on the left show the power spectrum of the vowel and the average firing rates of auditory nerve fibers as a function of their characteristic frequency. Average rate information from this number of fibers is not sufficient to identify the frequencies of individual harmonics in the stimulus that would be necessary to support a spectral pattern analysis of pitch. The plots on the right show the stimulus autocorrelation function and the population-interval distribution formed by summing all-order intervals from all fibers (i.e. the interval statistics of the whole auditory nerve). The population-interval distribution constructed from the responses of these fibers contains on the order of 100,000 intervals, and is sufficient to estimate the fundamental frequency (1/F0) with ~1% accuracy. This is in the ballpark of the ability of humans to detect changes in pitch on the order of fractions of a percent. A major advantage of interval codes over rate codes is that representations of stimulus periodicities remain highly invariant over a very wide dynamic range.

Neural coding and stimulus equivalence: an example of psychoneural isomorphism. Neural responses to four stimuli evoking a pitch at 160 Hz but differing in pitch salience. Left to right: Stimulus waveform, power spectrum, short term autocorrelation function, and population-interval distribution for each stimulus. Population-interval distributions are constructed by summing together the all-order interspike interval distributions of many auditory nerve fibers (n=49-85) having a wide range of characteristic frequencies. These interval distributions share a common feature: the most frequent interspike intervals present correspond to the pitch period and its multiples. While the top four stimuli give rise to strong definite pitches, the bottom two evoke weak, more diffuse pitches. Pitch salience qualitatively corresponds to the fraction of pitch-related intervals in the whole interval distribution (quantified in terms of peak-to-mean ratios in the distribution).

Temporal coding and speech perception
    Towards an temporal theory of speech perception
    Similarities and differences between autocorrelation and modulation detection operations
    Formant structure
    Voice pitch
    Voice onset time
    Temporal cues associated with frequency shifts
    Envelope dynamics
    Implications for cochlear implants
    Implications for speech recognition
        Interval-based front-ends for speech recognizers
        Running summary autocorrelations as alternatives to spectrograms
       Autocorrelation-based recognition strategies

Temporal coding and music perception
    Pitch and timbre
    Roughness
    Tonal fusion
    Harmonic relations
    Autocorrelation-based model of tonal context (e.g. Krumhansl note-key correlations)
    Rhythm (pdf working paper on timing nets and rhythm)
    Long-structure

Discussions of high level integration of neural information
Signal multiplexing, emergence of new signal-primitives, broadcast models of coordination (pdf, 204k)

Towards an evolutionary robotics:
a taxonomy of adaptive systems (Short paper, pdf )

    Purely computational devices
    Robotic devices with fixed sensors, effectors, and coordinative faculties
    Robotic devices with adaptive coordinative faculties
    Robotic devices with adaptive sensors and/or effectors
      • Gordon Pask's adaptive electrochemical device
    Self-constructing, self-modifying evolutionary robotic devices
    Epistemic autonomy: systems that determine their own categories vis-a-vis the environment

• Cariani P. On the design of devices with emergent semantic functions. Ph.D. Thesis, State University of New York at Binghamton, 1989; 234 pp. Download pdf file (7 Mb)

Theoretical biology
    • Links to other theoretical biology sites (forthcoming: Pattee, Rosen, Kampis, Conrad, Emmeche, Moreno & others)
    • Bibliography of theoretical biology
    • Interview with Robert Rosen on the evolution of his ideas
        (VHS-NTSC, 2 hrs, April, 1998, available upon request)

• Cariani, P. (1991) Emergence and artificial life. In: Artificial Life II, SFI Studies in the Sciences of Complexity, vol X, edited by C.G. Langton, C. Taylor, J.D. Farmer, & S. Rasmussen, Addison-Wesley, pp. 775-796. (Download pdf, 9 Mb)

Commentary on Cybernetics, Connectionism and Hayek's Sensory Order : CarianiOnSmith.html

Preserving alternative visions of the future: social institutions that enhance human dignity
The Mondragon industrial cooperatives: worker-owned, worker-controlled enterprises in a free-market framework
Gramin Banks: bottom-up economic development
Waiting for the Big One : 1) is the world financial system inherently stable?
2) long economic waves and dynamical systems, is Kondratieff really dead and gone?

Publications	Auditory neurophysiology	Neural timing nets	Adaptive systems	Epistemology Emergence	Semiotics
Cybernetics	Temporal codes	Music perception	Theoretical biology Artificial life	Philosophy of mind	Other topics