MLLO-IQ

• Information-based global optimization of f’, a quasi-Newton local optimization of f.

• Choose initial point x1

• Locally optimize x1 to x2, f’(x1)=f’(x2)=f(x2)

• Iterate:

  • Pick next point x1 according to g and previous f’ points
  • Locally optimize x1 to x2, f’(x1)=f’(x2)=f(x2)

• Complexity of g still grows with each iteration

• How to limit point taken into consideration?

Previous slide

Next slide

Back to first slide

View graphic version