Narahari Bharadwaj Songela Chen Chris Edgett Ishan Levy
Narahari Bharadwaj, Songela Chen, Chris Edgett, Ishan Levy, Agustey Mongia, Ruvini Navaratna, Martin Schneider, Claire Sheen, Nancy Wang, Justin Wong, Amy Zhou
What is global optimization? • What does it do? • How does it work? • Why use it? • Routing Problems • Biological Processes • Molecular Structure Prediction http: //www. mathworks. com/matlabcentral/fileexchange/scree nshots/4051/original. jpg
Would you like to try it? • Website allows users to query our solvers with custom input • Visual representations of the generated data • Unified theme for more effective presentation
How does the website work?
Traveling Salesman Problem (TSP) • No polynomial time algorithm exists What is the shortest path that traverses • every Focused city specifically in a graph once and returns to on Euclidean the beginning TSP (shortest Euclidean tour)?
TSP: Brute Force • Tests every possible route in order to find the shortest one • Has a runtime of O(n!) • n! ways to permute n distinct numbers
TSP: Ant Colony Optimization • Mimics ants’ pheromone system • Pheromone inversely related to total tour distance, determine the routes ants take • Final tour generated by always going along the route with the most pheromone
TSP: Fast/Wheel Dijkstra 1. Choose a random city 2. Find the closest city 3. Draw a route to the closest city 4. Repeat
TSP: Fast/Wheel Dijkstra
TSP: Fast/Wheel Dijkstra
TSP: Fast/Wheel Dijkstra
TSP: Gravitational • Works off of the idea that cities that are close to each other ‘attract’ and are favorable • Edges that connect two cities passing near the center of mass are considered unfavorable.
TSP: Gravitational
TSP: Gravitational
TSP: Simulated Annealing High Temperature "Redock: Docking Redone. " Redock: Docking Redone. N. p. , n. d. Web. 28 July 2014.
TSP: Simulated Annealing Medium Temperature "Redock: Docking Redone. " Redock: Docking Redone. N. p. , n. d. Web. 28 July 2014.
TSP: Simulated Annealing Low Temperature "Redock: Docking Redone. " Redock: Docking Redone. N. p. , n. d. Web. 28 July 2014.
TSP: Remove Line Intersections • Removes intersecting lines to reduce the length of the total path • Acts as a ‘helper’ algorithm • Resolve intersections by reversing order of nodes in between earliest and latest
Results for TSP
Results for TSP
Results for TSP
Results for TSP
Protein Folding What is the secondary structure for a certain arrangement of hydrophobic and hydrophilic amino acids?
Protein Folding: Slithering Snake • Generates a randomized chain of amino acids that travel like a snake around a coordinate plane • Calculates all possible configurations and returns the one with the lowest energy state http: //www. physics. buffalo. edu/phy 410505/2011/topic 5/app 1/img/culdesac. png
Protein Folding: Slithering Snake A more negative potential energy is favorable
Protein Folding: Neural Network • Forms configurations that correspond to hydrophobic and hydrophilic behavior
Results for Protein Folding
Results for Protein Folding
Results for Protein Folding
Nanoparticle Configuration Metallic nanoparticles show promise in many fields, but are they stable? How do we determine the configuration with the lowest potential energy?
Nanoparticle Configuration: Simulated Annealing "Redock: Docking Redone. " Redock: Docking Redone. N. p. , n. d. Web. 28 July 2014.
Nanoparticle Configuration: Genetic Algorithm • Create many possible particles and select ones with lowest energy states • These are ‘bred’ and ‘mutated’ for several generations
Nanoparticle Configuration: Basin Hopping 1. Randomly changes positions of atoms 2. Finds most stable state 3. Randomly makes smaller changes to atom positions D. J. Wales, "Frontiers Article: Surveying a Complex Potential Energy Landscape: Overcoming Broken Ergodicity Using Basin-Sampling, " Chem. Phys. Lett. 584: 1 -9 (2013).
Results for Nanoparticles
Results for Nanoparticles
Conclusion • Created a novel web interface for global optimization • Application: • Route Optimization • Drug Design • Fuel Cells
Acknowledgements Thank you to… • All the generous corporate and government sponsors who made PGSS possible • PGSS alumni for their passion in reviving PGSS • Zachary Pozun for his guidance and support • Jeff Conway for his unfailing help • Carnegie Mellon University for graciously providing educational and residential facilities • Dr. Barry Luokkala, faculty, and staff for making our PGSS experience unforgettable
- Slides: 39