Group 3. 


Good idea to visualize possible configurations of protein T. 


Critqiues: 

principle reason what Newton's won't work is not the derivative (which you
can actually get for the folding energy, I mentioned it in class). 

No exploration in 2D. Newton's is in 1D only. Newton's is not good for
global optimization. 

The analysis of the methods' performance is somewhat superficial. 


Group 5. 

Good idea to visualize possible configuration of protein T.

A well thought out "funnel" function. Clear algorithm for constructing it.

Methods timed.


Critiques: 

Exactly how % accuracy is defined? Unclear. 


Group 2. 

Good idea to visualize possible configuration of protein T.

Considered ASP and ARG pair. 

Good Convergence plots. 

Identified the key idea: don't get stuck in the local minimum. 


Critiques: 

Why use "random search" with inherently stochastic (random) methods?


Group 4. 

Considered ASP and ARG pair.

Timing.  

Critiques: 

The reason why CG or Newton's won't work is because it is a local method. 

No reason to emphasize "derivative-free" methods. 
It was mentioned in class that the energy function is differentiable. 


Group 1. 

Considered general advantages of various computational methods. 
Finished well within the 12 minute  time limit.  

Critiques: 

The proposed structures miss the key point: linear, no H-H contacts. 
That part was not worked out. 

No definition/exploration of your own folding funnel. 

Optimization discussion is completely off-traget. 
Why Euler method? Why Lagrange and Hermite polynomial? 
WE only need to discuss optimization methods.