Though the program is crowded there is still a sizeable crowd following the workshop, at 16 + \epsilon amounting to about 50 people.
Today's program has more long, 40 minute talks than yesterday (Thursday). We heard then from Riccardo Zecchina how to understand random Steiner trees using statistical mechanics (using a combination of belief propagation and the cavity equation approach) and then Jeremy Stribling discussed how to put together a distributed file system (or how to share data) "without problems": how to create a protocol for doing that in the P2P-spirit.
The Friday program has contained no less than five long talks. First Massimo Vergassola outlined the role of non-coding RNAs in bacteria. Then, Haijun Zhou overviewed his recent work on dynamics on networks: how the scale-free character influences opinion model dynamics (it is much dependent on the degree distribution exponent) and how such dynamics can be coupled with network evolution to create scale-free -looking networks. Scott Kirkpatrick told us how to treat the Sudoku puzzle as a challenging optimization problem (and how to solve it by various techniques). This was contrasted nicely by Ilkka Niemelä, whose presentation contained a lot of information on industrial SAT problems, how they are to be solved efficiently (and why), and what the real challenges in SAT are in the future. Finally, Neil Gershenfeld finished the day by a tale about how to program bits and atoms. Or, more precisely, how to inject programming into our surroundings, and how to learn from physics and biology to make such a holistic view work, including a mixture of nano-fabrication and really ubiquitous computing.