Organizational Design Optimization Using Genetic Programming Bijan - PowerPoint PPT Presentation

Organizational Design Optimization Using Genetic Programming Bijan KHosraviani, Raymond E. Levitt and John R. Koza S tanford University GECCO 2004 Conference June 30, 2004

Presentation Outline Conclusions � Introduction / Motivation Methodology Results � Objectives / Research Questions � Research Methodology � Results to Date Objectives � Conclusions Introduction Introduction Intro June 30, 2004 Organizational Design Optimization 2

Evolution of Organization Design Conclusions Trial-&-Error Org’n Analysis: Org’n Design: Adaptation VDT/SimVision VDT+Optimizer Methodology Results 1. Set project objectives. 1. Set project objectives. 1. Set project objectives. 2. Propose organization. 2. Propose alternative 2. Propose initial organizations. organization as starting point for optimization. 3. Complete project 3. Model alternative 3. Evolve many alternative using proposed organizations and organizations; predict Objectives organization and simulate each one to performance of each observe outcome. predict outcomes. one; evaluate “fitness”. 4. Succeed or fail. 4. Choose solution that 4. Evolve optimal organi- Try to learn and optimizes outcomes. zational configuration by adapt. selective reproduction & Introduction Introduction mutation of alternatives. Intro June 30, 2004 Organizational Design Optimization 3

Motivation Conclusions � Project organization design is a complex, multi- 1 Project Lead dimensional, optimization Management Methodology Results Meeting problem 1 1 1 1 1 Foundry Team Test Marketing Team 1 Chip Architect Foundry Lead Meeting Engineering St Architecture Team 1 1 Meeting 0.8 1 1 � Analysis tools exist for Logic Design Verification Foundry Layout Foundry Test Team 1 Team Engineer Engineer organizational design, but 1 Design 1 no known automated 1 Coordination 1 1 optimizer exists 4 1 4 1 Objectives 4 1 1 Gen Test Suite 4 1 1 Develop Fab Test Generate Test Start Specification & Deliver Vectors � Finding an optimal or near- Insert Scan Write B1RTL Verify RTL Partition Chip optimal solution is a Verify B1RTL FullChipSynth Sim Gates Assemble RTL challenging task even for Introduction Introduction Synth_B1RTL an experienced PM FloorPlanning PlaceRoute PhysVerifn Intro June 30, 2004 Organizational Design Optimization 4

Objectives / Research Questions Conclusions Objectives � Develop an optimizer for VDT using evolutionary computing Methodology Results techniques to help project managers find near-optimal designs for their project organizations � Validate the postprocessor against both theory and practice Research Questions � How can GP help a highly experienced manager in designing Objectives a project organization? Objectives � Are “optimal” solutions found by GP in-line with organization theory and management best practices? � What are the limits of GA/GP for organization design? Introduction Introduction Intro June 30, 2004 Organizational Design Optimization 5