1.1 --- a/org/ai-journal-review.org	Tue Apr 16 13:59:06 2013 +0000
     1.2 +++ b/org/ai-journal-review.org	Sun Apr 21 17:05:30 2013 +0000
     1.3 @@ -1,12 +1,37 @@
     1.4  #+title:Interesting Papers in Artificial Intelligence
     1.5  
     1.6  I decided to read all of the /titles/ in the Artificial Intelligence
     1.7 -journal, and found these interesting papers.  The entire process took
     1.8 -about 2 hours.
     1.9 +journal, and found these interesting papers.  The entire title-reading
    1.10 +process took about 2 hours.
    1.11  
    1.12  * Interesting Concept
    1.13  
    1.14 -Jordi Delgado - Emergence of social conventions in complex networks
    1.15 +- (2002) Jordi Delgado - Emergence of social conventions in complex networks
    1.16 + 
    1.17 +  Here, "social conventions" means a very specific property of graphs
    1.18 +  in the context of game theory. Their social networks are groups of
    1.19 +  mindless automotaons which each have a single opinion that can take
    1.20 +  the values "A" or "B". They use the "coordination game" payoff
    1.21 +  matrix that engourages each pair of agents to agree with each other,
    1.22 +  and study various ways the graph can come to 90% of the agents all
    1.23 +  believe either "A" or "B".  It's probably not useful for actual
    1.24 +  social worlds, and there's no simulation of any interesting
    1.25 +  environment, but it might be useful for designing protocols, or as a
    1.26 +  problem solving method.
    1.27 +
    1.28 +  References: 
    1.29 +  + L.A. Nunes Amaral, A. Scala, M. Barthélémy, H.E. Stanley, Classes
    1.30 +    of small-world networks, Proc. Nat. Acad. Sci. 97 (2000)
    1.31 +    11149–11152.
    1.32 +  + D.J Watts, S.H. Strogatz, Collective dynamics of small-world
    1.33 +    networks, Nature 393 (1998) 440–442.
    1.34 +  + Y. Shoham, M. Tennenholtz, On the emergence of social conventions:
    1.35 +    Modeling, analysis and simulations, Artificial Intelligence 94
    1.36 +    (1997) 139–166.
    1.37 +
    1.38 +- (1997) Yoav Shoham, Moshe Tennenholtz - On the emergence of social
    1.39 +  conventions: modeling, analysis, and simulations
    1.40 +
    1.41  
    1.42  Marcelo A. Falappa, Gabriele Kern-Isberner, Guillermo R. Simari -
    1.43  Explanations, belief revision and defeasible reasoning
    1.44 @@ -66,8 +91,8 @@
    1.45  Paul Snow - The vulnerability of the transferable belief model to
    1.46  Dutch books
    1.47  
    1.48 -Simon Kasif, Steven Salzberg, David Waltz, John Rachlin, David W. Aha
    1.49 - - A probabilistic framework for memory-based reasoning
    1.50 +Simon Kasif, Steven Salzberg, David Waltz, John Rachlin, David
    1.51 +W. Aha - A probabilistic framework for memory-based reasoning
    1.52  
    1.53  Geoffrey LaForte, Patrick J. Hayes, Kenneth M. Ford - Why Gödel's
    1.54  theorem cannot refute computationalism
    1.55 @@ -95,9 +120,6 @@
    1.56  Shmuel Onn, Moshe Tennenholtz - Determination of social laws for
    1.57  multi-agent mobilization
    1.58  
    1.59 -Yoav Shoham, Moshe Tennenholtz - On the emergence of social
    1.60 -conventions: modeling, analysis, and simulations
    1.61 -
    1.62  Stuart J. Russell - Rationality and intelligence
    1.63  
    1.64  Hidde de Jong, Arie Rip - The computer revolution in science: steps
    1.65 @@ -151,9 +173,6 @@
    1.66  
    1.67  Takeo Kanade - From a real chair to a negative chair
    1.68  
    1.69 -Berthold K.P. Horn, B.G. Schunck - “Determining optical flow”: a
    1.70 -retrospective
    1.71 -
    1.72  Harry G. Barrow, J.M. Tenenbaum - Retrospective on “Interpreting line
    1.73  drawings as three-dimensional surfaces”
    1.74  
    1.75 @@ -597,7 +616,61 @@
    1.76  Larry S. Davis, Azriel Rosenfeld - Cooperating processes for low-level
    1.77  vision: A survey
    1.78  
    1.79 -Berthold K.P. Horn, Brian G. Schunck - Determining optical flow
    1.80 +- (1980) Berthold K.P. Horn, Brian G. Schunck - Determining optical
    1.81 +  flow
    1.82 +
    1.83 +  Optical flow is an estimation of the movement of brightness
    1.84 +  patterns. If the image is "smooth" then optical flow is also an
    1.85 +  estimate of the movement of objects in the image (projected onto the
    1.86 +  plane of the image). They get some fairly good results on some very
    1.87 +  contrived examples. Important point is that calculating optical flow
    1.88 +  involves a relaxation process where the velocities of regions of
    1.89 +  constant brightness are inferred from the velocities of the edges of
    1.90 +  those regions.
    1.91 +  
    1.92 +  This paper is a lead up to Horn's book, Robot Vision.
    1.93 +
    1.94 +  Hexagonal sampling may be a good alternative to rectangular
    1.95 +  sampling.
    1.96 +
    1.97 +  A reduced version of this algorithm is implemented in hardware in
    1.98 +  optical mice to great effect.
    1.99 +
   1.100 +  + Hamming, R.W., Numerical Methods for Scientists and Engineers
   1.101 +    (McGraw-Hill, New York, 1962).
   1.102 +  + Limb, J.O. and Murphy, J.A., Estimating the velocity of moving
   1.103 +    images in television signals, Computer Graphics and Image
   1.104 +    Processing 4 (4) (1975) 311-327.
   1.105 +  + Mersereau, R.M., The processing of hexagonally sampled
   1.106 +    two-dimensional signals, Proc. of the IEEE 67 (6) (1979) 930-949.
   1.107 +
   1.108 +
   1.109 +- (1993) Berthold K.P. Horn, B.G. Schunck - “Determining optical flow”: a
   1.110 +  retrospective
   1.111 +  
   1.112 +  Very useful read where Horn criticies his previous paper.
   1.113 +
   1.114 +  - Whishes that he distinguished "optical flow" form "motion
   1.115 +    field". "Optical flow" is an image property, whilc the "motion
   1.116 +    field" is the movement of objects in 3D space. "Optical flow" is a
   1.117 +    2D vector field; the "motion field" is 3D.
   1.118 +  - Wished he made the limitations of his algorithm more clear.
   1.119 +  - His original paper didn't concern itself with flow segmentation,
   1.120 +    which is required to interpret real world images with objects and
   1.121 +    a background.
   1.122 +  - Thinks that the best thing about the original paper is that it
   1.123 +    introduced variational calculus methods into computer vision.
   1.124 +    
   1.125 +  References:
   1.126 +
   1.127 +  + R. Courant and D. Hilbert, Methods of Mathematical Physics
   1.128 +    (Interscience, New York, 1937/1953).
   1.129 +  + D. Mart, Vision (Freeman, San Francisco, CA, 1982).
   1.130 +  + C.M. Thompson, Robust photo-topography by fusing
   1.131 +    shape-from-shading and stereo,Ph.D. Thesis, Mechanical Engineering
   1.132 +    Department, MIT, Cambridge, MA (1993).
   1.133 +  + K. Ikeuchi and B.K.P. Horn, Numerical shape from shading and
   1.134 +    occluding boundaries, Artif lntell. 17 (1981) 141-184.
   1.135  
   1.136  Katsushi Ikeuchi, Berthold K.P. Horn - Numerical shape from shading
   1.137  and occluding boundaries
   1.138 @@ -617,6 +690,42 @@
   1.139  
   1.140  * Cryo!
   1.141  
   1.142 -Kenneth D. Forbus, Peter B. Whalley, John O. Everett, Leo Ureel, Mike
   1.143 -Brokowski, Julie Baher, Sven E. Kuehne - CyclePad: An articulate
   1.144 -virtual laboratory for engineering thermodynamics
   1.145 +- (1999) Kenneth D. Forbus, Peter B. Whalley, John O. Everett, Leo
   1.146 +  Ureel, Mike Brokowski, Julie Baher, Sven E. Kuehne - CyclePad: An
   1.147 +  articulate virtual laboratory for engineering thermodynamics
   1.148 +     
   1.149 +  Should learn about thermodynamics, and about "thermal cycles."
   1.150 +  http://www.qrg.northwestern.edu/projects/NSF/cyclepad/cyclepad.htm
   1.151 +  
   1.152 +  This system is more about expressing models and assumtions than
   1.153 +  automatically generating them, and as such is similiar to our "math
   1.154 +  language" idea.
   1.155 +
   1.156 +  It's like a simple circuit modeller, and similar to Dylan's idea of
   1.157 +  an online circuit modeler.
   1.158 +  
   1.159 +  #+begin_quote
   1.160 +  We found that if CyclePad did not do the “obvious” propagation in
   1.161 +  preference to interpolation, students trusted it less.
   1.162 +  #+end_quote
   1.163 +  
   1.164 +  It's too bad that the paper doesn't mention the shortcommings of the
   1.165 +  system. 
   1.166 +  
   1.167 +  + J.O. Everett, Topological inference of teleology: Deriving    
   1.168 +    function from structure via evidential reasoning, Artificial
   1.169 +    Intelligence 113 (1999) 149–202.
   1.170 +  + P. Hayes, Naive physics 1: Ontology for liquids, in: J. Hobbs,
   1.171 +    R. Moore (Eds.), Formal Theories of the Commonsense World, Ablex,
   1.172 +    Norwood, NJ, 1985.
   1.173 +  + P. Nayak, Automated modeling of physical systems, Ph.D. Thesis,
   1.174 +    Computer Science Department, Stanford University, 1992.
   1.175 +  + R.W. Haywood, Analysis of Engineering Cycles: Power, Refrigerating
   1.176 +    and Gas Liquefaction Plant, Pergamon Press, 1985.
   1.177 +  + R.M. Stallman, G.J. Sussman, Forward reasoning and
   1.178 +    dependency-directed backtracking in a system for computer-aided
   1.179 +    circuit analysis, Artificial Intelligence 9 (1977) 135–196.
   1.180 +    + Dylan should read this, since it concerns his online circuit
   1.181 +      analysis idea.
   1.182 +
   1.183 +