Fereshteh Shahmiri

PhD in Computer Science - School of Interactive Computing - Georgia Institute of Technology

 

Technology Square Research Building
85 Fifth Street NW
Atlanta, GA 30308

© 2017 by Fereshteh Shahmiri. All rights reserved.

 

Carnegie Mellon University |Machine Learning Department, School of Computer Science

 

Course: Algorithms in Nature

Special Thanks to Professor Ziv Bar Joseph

http://www.cs.cmu.edu/~zivbj/

Fall 2013

 

 

 

Research - Project 1 |

Researched on design solutions optimized for energy efficiency

 

Title: MAXIMIZING PRESERVED SUNLIGHT IN PNEUMATIC SOLAR PANELS

            APPLICATION | Building Facade Components

            APPROACH | OPTIMIZATION - GENETIC ALGORITHM

 

Fereshteh Shahmiri, Alan Shteyman

Paper Link Here
 

 

 

Research - Project 2 |

Researched on one important part of DATA MINING PROBLEMS | FINDING DENSE MODULES OR CLUSTERS IN A GRAPH

 

Title: GENETIC ALGORITHM TO CLUSTER GRAPHS

 

Fereshteh Shahmiri

Project and Code Implementation Link Here

 

DESCRIPTION |

Finding dense module or clusters in a graph is an important part of many data mining problems. One popular definition of a 'module' is a set of nodes that have many more within-module connections (i.e. connections between nodes in the same modules) than expected by chance. In 2002, Newman proposed an objective function, called modularity, that characterized the quality of clustering C of a graph G = ( V, E)

The Goal is to find the clustering C that MAXIMIZE Newman Function. In general the clustering C can have any number of modules( from 1 to n, where n is the number of nodes in the graph) but all nodes must be assigned to exactly one module.

 

PROJECT DEFINITION |

Writing a genetic program to cluster an input graph into modules that optimized the Newman objective function, I used here at most 5 clusters.

 

EXPECTED OUTPUT |

Program should output the OPTIMAL MODULARITY found as well as the CLUSTERING corresponding to the optimal modularity.

 

POINTS NEED TO BE CONSIDERED |

  • For this problem, a CLUSTERING is the analog of an 'individual'.

  • The 'FITNESS FUNCTION' is equivalent to the 'MODULARITY FUNCTION'.

  • The main idea behind the solution and using the operations like MUTATION and CROSS-OVER, etc. and the way of ensuring that how these operation always produce a valid clustering. (e.g. how it is ensured that a node was not assigned to multiple clusters?)

 

 

 

RELEVANT RESEARCH ON GENETIC ALGORITHM |

 

Part 1 |

PROBABILITY OF PRESERVING THE SCHEMA  FROM DESTRUCTION OCCURING BY SOME OPERATIONS LIKE MUTATION

 

DESCRIPTION | In genetic algorithms, schema is a template that defines a set of possible strings. A String destroys the schema if the string is not a possible string defined by the schema.Research structured on understanding how the probability that a schema will not be destroyed by operations like mutation that occurs with specific number for its probability, is calculated?

 

Part 2 |

CALCULATION OF AVERAGE FITNESS OF THE EXPLAINED SCHEMA

 

 

 

RESEARCH - PROJECT 3 |

Based on specific given data, how first two principal components that would be found using the PCA algorithm have to be drawn?

How is the reconstruction error if the choice is ONE component? Or TWO components?

Title : DIMENSIONALITY REDUCTION

 

 

LECTURE TOPICS:

 

  • Introduction to biology

  • Introduction to distributed computing

  • Regression

  • Neural Networks

  • Dimensionality Reduction

  • Non-negative matrix factorization

  • Optimization and Search

  • Genetic Algorithms

  • Ant Algorithms

  • Maximum Independent Set (MIS)

  • Lateral inhibition

  • MIS and SOP

  • Steiner Trees

  • Slime Mold Design

  • Network Growth Models

  • Pruning Algorithms

  • Robustness