This week's finds in the journals

Anders Sandberg (
13 Feb 1999 00:59:53 +0100

	Making myoglobin better at breaking down peroxides
	Spinal microstimulation for erection
	An unusual memory enhancer
	No free lunch theorems for optimization
	Glyphs for visualizing data	
	Information murals
	Startlecam: a wearable camera linked to attention
	Using small-world networks
	Quantum game theory

In vitro evolution of horse heart myoglobin to increase peroxidase activity L. Wan, M. B. Twitchett, L. D. Eltis, A. G. Mauk and M. Smith Proc Natl Acad Sci U S A, 95:22 12825--31 Oct 27 1998

Myoglobin, just like hemoglobin, is able to break down hydrogen peroxide. In this experiment they made a gene for the protein, subjected it to mutations, implanted it in bacteria, screened for peroxidase activity, took the best version and continued the process. The result was a protein 25 times as active as the original. Nice to see that it is possible to improve on nature using so few iterations.

Penile erection produced by microstimulation of the sacral spinal cord of the cat
C. Tai and A. M. Booth and W. C. de Groat and J. R. Roppolo IEEE Trans Rehabil Eng, 6:4 374--81, Dec 1998

The neural pathways of erection are rather messy, but this paper studies how microstimulation in the spinal cord can cause it. The obvious application is of course for helping people with spinal cord injury, but I think there might be transhumanist potential for hedonic engineering here.

Phlorizin, a competitive inhibitor of glucose transport, facilitates memory storage in mice
M. M. Boccia, S. R. Kopf and C. M. Baratti Neurobiol Learn Mem, 71:1 104-12, Jan 1999

And the obligatory new memory enhancer :-) It is known that glucose enhances memory encoding, so it would appear that an inhibitor like phlorizin would make encoding worse, but the opposite is true. When injected after training it made mice quicker at avoiding an electric shock they had learned before, but there was no effect if given just before the testing, so it just enhances encoding, not retrieval. It shows the usual inverted U-curve, and counteracted the memory decrease caused by insulin. So it seems that it is enough that a substance is glucose-like to affect memory, very odd.

No Free Lunch Theorems for Optimization
David H Wolpert and Wiliam G Macready
IEEE Transactions on Evolutionary Computation, 1:1 67--82 1997

It might look like some ways of optimizing are better than others (like genetic algorithms being better than hillclimbing which is better than random search), but surprisingly it is wrong. If an algorithm for finding an optimim performs better than another algorithm on a certain set of optimization problems, then the reverse must be true for all other optimization problems! This means that there is no best method, and in fact they are all just as bad as random search. This might look very strange or even depressing, but it is really fairly trivial: the set of all optimization problems include a vast majority of completely intractable problems (like finding the optimum of a pseudorandom function), the set of "reasonable" problems we are used to is vanishingly small. Over a certain set of problems, there are certainly functions that are better than others. This is just another demonstration that prior knowledge is essential for problem solving.

Information Rich Glyphs for Software Management Data Mei C Chuah and Stephen G Eick
IEEE Computer Graphics and Applications, 24--29 July/August 1998

Just some more visualization tricks, this time glyphs representing things such as added lines, errors, number of changes, people etc in a database of software releases. They use a wheel-like or bug-like glyphs and discuss their advantages and disadvantages. Maybe useful for analyzing mailing list archives?

The Information Mural: A Technique for Displaying and Navigating Large Information Spaces
Dean F Jerding and John T. Stasko IEEE Transactions on Visualization and Computer Graphics, 4:3 257--271, July-September 1998

Another nice way of visualizing data: have a shrunken "mural" depicting the whole document/database at the side of the magnified view you watch. Makes it possible to both get an overview, notice patterns and search.

StartleCam: a cybernetic wearable camera Healey, J.; Picard, R.W.
Wearable Computers, 1998. Digest of Papers. Second International Symposium on

A fun system, now actually implemented in reality. The wearer has sensors measuring galvanic skin response, and a wearable video camera. If the skin response rises, the system stores the pictures leading up to that point. Seems to be able to trigger not just to surprises or stress, but also attention. Very interesting, and suggests that affective computing can become quite useful.

Characterization and control of small-world networks S. A. Pandit, R. E. Amritkar

Small-world networks (the kind of networks where some long-range connections makes almost every node close to every other node, despite having mostly local connections; examples include collaboration graphs, neural networks and power grids) appears to be useful for a lot of things. This paper studies how the sensitive dependence on the long-range connections makes it possible to control the spread of an epidemic by first immunizing the long-range connections. It also turns the problem around, and sufggests that the best way of spreading advertisements (or memes) would be to target the long-range links.

Quantum Strategies
David A. Meyer Physical Review Letters 82:5 1052--1055 1999

Since computation can be generalized to quantum computation, why not generalize game theory? This paper deals with the odd things that can happen if you allow players to choose quantum superpositions of strategies. As an example Meyer describes a game between captain Picard of Star Trek and the superbeing Q, where they take turns of either flipping a hidden coin or not. Classically, this is a fair zero-sum game with an equilibrium of mixed strategies. But if Q makes quantum moves, then he can win against the classical Picard! There is no equilibrium if both players play deterministic quantum strategies, but Meyer shows that there is an equilibrium if both plays mixed quantum strategies (i.e. they select randomly between a number of quantum strategies). A fun generalization, and it ties in with quantum error correction (where nature is one of the players).

Anders Sandberg                                      Towards Ascension!                  
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y