Dear Loet, Steven, and colleagues,

During last ten years or so, with particular success in most recent years, Karl Friston has developed his free energy optimization principle, based on Shannon's information theory and optimal control theory as well as on the Bayesian brain hypothesis. I think this is the most advanced work towards a unified brain theory today. The minimization dynamics of the cerebral free energy construct (it is a sort of Helmoltz program revisited) becomes a generative process of perception, action, learning and adaptive behaviors in general. The 2010 paper (Nature Reviews Neurosceince, doi: 10.138/nrn2787) where he precisely argues about a unified brain theory, is quite representative of his proposals. On a personal basis, during last two decades I was following and cooperating with Kenneth Paul Collins (we published a book in Spanish about the emergence of behavior from brain dynamics). Our scheme was based on the minimization of a collective variable supposedly a sort of "entropy" of excitation/inhibition ratios topologically distributed among neuronal surfaces of the cortex that was performed essentially by the medial parts of the brain. Although very rich in qualitative and behavioral aspects, the formal part was too weak (awfully weak). Until recent years I could not connect meaningfully Collin's approach with other works, and unfortunately he left scientific research long ago--but now the marriage with Friston's is remarkable. Putting them together may be a very fertile exploratory avenue.

best ---Pedro

Loet Leydesdorff wrote:

Dear Steven and colleagues,

I did not (yet) study your approach. Is there a paper that can be read as an introduction?

It seems to me that one can distinguish between formal and substantial theories of information. Shannon’s mathematical theory is a formal apparatus: the design and the results do not yet have meaning without an interpretation in a substantial context. On the other side, a theory about, for example, neuro-information is a special theory. One can in this context use information theory as a statistical tool (among other tools). Sometimes, one can move beyond description. J

The advantage of information theory, from this perspective of special theories, is that the formal apparatus allows us sometimes to move between domains heuristically. For example, a model of the brain can perhaps be used metaphorically for culture or the economy (or vice versa). The advantages have to be shown in empirical research: which questions can be addressed and which puzzles be solved?




Loet Leydesdorff

/Emeritus/ University of Amsterdam
Amsterdam School of Communications Research (ASCoR) <>; Honorary Professor, SPRU, <>University of Sussex;

Guest Professor Zhejiang Univ. <>, Hangzhou; Visiting Professor, ISTIC, <>Beijing;

Visiting Professor, Birkbeck <>, University of London; <>

*From:* [] *On Behalf Of *Steven Ericsson-Zenith
*Sent:* Tuesday, December 09, 2014 10:13 PM
*Cc:* Joseph Brenner; fis
*Subject:* Re: [Fis] Information-as-Process

The problem with this approach (and approaches like it) is that it is descriptive and not explanatory. The distribution of the shape, in my model, can be described, perhaps, but the process or action decision point and response covariance is impossible to consider. It is for this reason that I use holomorphic functors and hyper-functors in which I can express the explicit role of a base universal (per gravitation).

Nor is it clear to me that this is what Joe referred to as "information as process."

On Mon, Dec 8, 2014 at 10:20 PM, Loet Leydesdorff < <>> wrote:

    Dear colleagues,

    Shannon’s information theory can be considered as a calculus
    because it allows for the dynamic extension. Theil
    (1972)—Statistical decomposition analysis (North
    Holland)—distinguished between static and dynamic information
    measures. In addition to Shannon’s statical H, one can write:


    in which
    be considered as the a posteriori and
    a priori distribution. This dynamic information measure can be
    decomposed and aggregated. One can also develop measures for
    systemic developments and critical transitions. In other words,
    information as a process can also be measured in bits of
    information. Of course, one can extend the dimensionality (/i/)
    for the multivariate case (/ijk/…), and thus use information
    theory for network analysis (including time).




    ·        Leydesdorff, L. (1991). The Static and Dynamic Analysis
    of Network Data Using Information Theory. /Social Networks,
    13/(4), 301-345.

    ·        Theil, H. (1972). /Statistical Decomposition Analysis/.
    Amsterdam/ London: North-Holland.


    Loet Leydesdorff

    /Emeritus/ University of Amsterdam
    Amsterdam School of Communications Research (ASCoR) <>;
    Honorary Professor, SPRU,
    <>University of Sussex;

    Guest Professor Zhejiang Univ. <>,
    Hangzhou; Visiting Professor, ISTIC,

    Visiting Professor, Birkbeck <>, University
    of London;

    *From:* Fis [
    <>] *On Behalf Of *Steven
    *Sent:* Monday, December 08, 2014 10:22 PM
    *To:* Joseph Brenner
    *Cc:* fis
    *Subject:* Re: [Fis] Information-as-Process

    I am a little mystified by your assertion of "information as
    process." What, exactly, is this and how does it differ fro
    information in general (Shannon). Is it related to Whitehead's
    process notions?

    In terms of neuroscience it is important to move away from
    connectionism and modern computational ideas I believe. It is not
    clear to me how information theory can be applied to the operation
    of the brain at the synaptic level because the actions and the
    decisions made are made across the structure and not at a single
    Recognition, for example, is not a point event but occurs rather
    when a particular shape is formed in the structure (of the CNS,
    for example) and is immediately covariant with the "appropriate"
    response (another shape) which may be characterized as a
    hyper-functor (which may or may not include neurons and astrocytes
    in the brain).



Pedro C. Marijuán
Grupo de Bioinformación / Bioinformation Group
Instituto Aragonés de Ciencias de la Salud
Centro de Investigación Biomédica de Aragón (CIBA)
Avda. San Juan Bosco, 13, planta X
50009 Zaragoza, Spain
Tfno. +34 976 71 3526 (& 6818)

Fis mailing list

Reply via email to