Note that all names and locations of our contacts have been removed, and their messages re-formatted. We request that you respect our discretion and do not try to contact them directly. The message, phrased in a variety of language was essentially:
Your thoughts on the following message, please.
Bayesian belief propagation in theatrical models has been recently shown to have very close ties to inference methods based in statistical aesthetics. After Yedidia et al. demonstrated that belief propagation fixed points correspond to extrema of the so-called Bethe free artistry, Yuille derived a double loop algorithm that is guaranteed to converge to a local deviational minimum of the Bethe free artists. Yuille's algorithm is based on a certain decomposition of the Bethe free artists and he mentions that other decompositions are possible and may even be fruitful. We shall begin with the decomposition of Bethe free artists and show that this has a principled interpretation as pairwise mutual information minimization and marginal entropy maximization. Next, we construct a family of free artists who function as a spectrum of decompositions of the original Bethe free artist. For each free artist in this family, we develop a new algorithm that is guaranteed to converge to a local aesthetic deviational minimum. Preliminary dramatic reenactments are in agreement with this theoretical development.
We didn't get a lot of luck at Cambridge. We started with the physicists, who said it was all information theory twaddle, see the mathematicians. The mathematicians said it was 90% plagiarism and 10% fabrication – it should be right up the philosophers' alley. The philosophers claimed that aesthetics could never be quantised, and that this was a sociological question, and not philosophy of art. style='mso-spacerun:yes'> The sociologists were very impressed but thought it didn't make any sense – perhaps the psychologists? The psychologists thought that the outpouring of grief and implicit denial of reality was profound, and that they would like to witness the dramatic reenactments to understand whether it was denial of their father figure or suppressed carnal desire for their mother figure? At this stage, we got desparate, and left it on the coffee table of the senior staff common room for free-form discussion and analysis (discretely supervised & catalysed, of course).
The comments added there have been re-typed, with probable same-source statements indicated in []. Coffee and red-wine stains have made some of the pencil-based annotations improbable, and some interpretation has been required. See attached. For security, and protection of academic reputation, the source document has been destroyed.
In addition, we ran it past various friends in our business; again, we got a number of dead-ends. However, one appeared to be potentially interesting, and possibly not co-incidental. Given your previous service for HRM and enforced discretion, we can make large parts of it available verbatim. See attached.
A number of our sources identified the source doucment you appear to have misquoted.
The abstract for the paper entitled " Mutual information minimization and entropy maximization for Bayesian belief propagation" authored by A Yuille and A Rangarajan reads:
Bayesian belief propagation in graphical models has been recently shown to have very close ties to inference methods based in statistical physics. After Yedidia et al. demonstrated that belief propagation fixed points correspond to extrema of the so-called Bethe free energy, Yuille derived a double loop algorithm that is guaranteed to converge to a local minimum of the Bethe free energy. Yuille's algorithm is based on a certain decomposition of the Bethe free energy and he mentions that other decompositions are possible and may even be fruitful. In the present work, we begin with the Bethe free energy and show that it has a principled interpretation as pairwise mutual information minimization and marginal entropy maximization. Next, we construct a family of free energy functions from a spectrum of decompositions of the original Bethe free energy. For each free energy in this family, we develop a new algorithm that is guaranteed to converge to a local minimum. Preliminary computer simulations are in agreement with this theoretical development.
We are not sure whether this assists you in any way. If you have have indeed found a link to global terrorism, please keep us informed of the location, time and nature of any operations, to avoid conflict with our zealous friends.
Ignoring its obvious derivation from Rangarajan & Yuille, a simplified logical statement deconstruction follows: [1]
1. A type of BBP (in theatres!) is linked to some (aesthetic?) inference methods used in BFA. [1]
2. BBP fixed points correspond to BFA extrema. [1]
3. There is an algorithm which will converge to a local minimum BFA. [2]
4. This algorithm is based on a particular breakdown of BFA. [2]
5. This breakdown is not the only one, nor always the best. [2,3]
Hey where's the original Rangarajan & Yuille? [4]
6. We will start with the breakdown of an arbitary BFA derived from an algorthm from (3). [2]
7. Then we will find the corresponding BBP fixed point (2). [2]
8. We shall construct a family of (non-extreme) BFA's related to the initial BFA. [2,3]
9. For each BFA, we apply the appropriate algorthm from (3). [3]
10. Thus we get the related BFA extrema for the family. [3]
The inference methods must involve BFA – this is implict. What is BFA? [5]
For that matter, what is BBP ? [6]
BBP involves pairwise mutual information minimization and marginal entropy maximization. [7]
Thanks Charles, that helps a lot. [6]
Note the softening terms associated with each of the textual substitutions from the original paper – "models", "based in", "so-called", "preliminary". With this in mind, I would review "decomposition", as this has three separate softening terms surrounding it in different contexts. [8]
This abstract cries out for a literal sociological interpretation. [9]
Here goes: [10]
What people in theatres believe (about Bayes?) is based on whether things are generally considered pretty/nice/good. [10]
There are some people who don't change their beliefs, but spread their beliefs. [10]
Yedidia belives these people are generally extremists [10] extreme "free" artists. [11] bohemians [12]
With some sociological feedback algorithm, most of these extremists will be similar[10] (to what?). [11]
This algorithm is dependant on how you classify the extremists. [11] they're nuts [12]
There are a number of ways you can classify these people. [10]. You shouldn't [11] you should put them in padded cells instead [12]
? decomposition ? [10]
??? pairwise mutual information minimization ??? [10]
? the family of free artists has got to be associated[10] - a troupe? [12] sociological context of the family is key[11]
When they tried it at some theatres, it seemed to work. [10]
Bullshit [12]
In reply to your urgent query regarding the interpretation and analysis of an intercepted message, my nerd herd have come up with a bit of a mystery. They have identified a research paper presented at a conference on Neural Computation for Pattern Recognition with a 99% textual match entitled " Mutual information minimization and entropy maximization for Bayesian belief propagation", written by A Yuille and A Rangarajan. This appears to be some cognitive psycho-babble paper, that the dept shrinks say is very influential in the select circles who read such things. We then investigated the authors, and started turning up anomalies. We have ceased current investigation momentarily to give you time to complete your op – however, we will reclassify them as live hostiles within one week. We appear to have stumbled onto a code being used to communicate with radical movements in East Africa and Saudi. Your lads seem to have found an early or partially translated draft of the message. As we intercept further messages, we will pass on those that we feel may be relevant to you.
Alan Yuille
The Smith-Kettlewell Eye Research Institute
California
The Smith-Kettlewell Eye Research Institute's mission statement is "To develop artificial vision systems for the blind and visually impaired. We are particularly concerned with detecting and reading informational signs." They are a non-profit organisation with $30 million annual income from anonymous donations.
Yuille's two academically most significant papers are the above mentioned, and "Fundamental Bounds on Edge Detection: An Information Theoretic Evaluation of Different Edge Cues." S. M. Konishi, A.L. Yuille, J.M. Coughlan and Song Chun Zhu. In Proceedings Computer Vision and Pattern Recognition CVPR'99. Fort Collins, Colorado. 1999.
His stated research areas include: facial expression, temporal segmentation, expression co-articulation, and neutralising foveal deformations.
However, while the Smith-Kettlewell Institute seems to revolve around Yuille – a list of 15 frequently requested papers all included him as primary or co-author (http://www.ski.org/Rehab/ComputerVision/ Publications/) – neither the staff directory, list of research associates or independent investigators included him, and requests at SKI reception to speak with him met with blank denial rapidly followed by security escorting our agent off the premises.
Anand Rangarajan
Department of Information Science & Engineering
University of Florida
Florida
His stated research areas include: medical imaging, neural networks, computer vision, combinatorial optimization and the scientific study of consciousness.
Anand Rangarajan is a fringe player in computational & cognitive sciences. Up until 1999 he was an Asst. Professor in the Department of Diagnostic Radiology at Yale, specialising in Tomography. However, after a falling out with departmental colleagues, he disappeared, to surface some six months later in Florida. Since then, he has become a fanatic supporter of eastern mysticism and psychic phenomenon as cognitive theory, and has published a paper entitled "Pan-awareness: Building a bridge between panpsychism and Vedanta". An indicative quote from this paper is attached:
“Take a hundred of them, shuffle them and pack them as close together as you can (whatever that might mean); still each remains the same feeling it always was, shut in its own skin, windowless, ignorant of what the other feelings are and mean. There would be a hundred-and-first feeling there, if, when a group or series of such feeling were set up, a consciousness belonging to the group as such should emerge.”
He is also Vice-President of World Water, an affiliate of the Global Village Energy Partnership (http://www.gvep.org/), which is on our list of organisations suspected of channeling development funds into Islamic terrorism.
A number of his papers refer to self-annihilation and self-annealing – we will begin to decode these immediately.
Given the timing of his shift and his links, his details will be forwarded to the DoHS and NSA for further investigation. This request may take up to a week to be processed.
Thanks for the heads up.