Algorithmic differentiation and the optimization strategy in TurboRVB

Topical course given by Sandro Sorella

Start: 02-04-2018 End: 20-04-2018 Room: 131
Schedule: Tue 11:00 - 13:30, Wed 14:30 - 17:00

Program

  • Algorithmic differentiation : the idea
  • Adjoint algorithmic differentiation: how to make efficient algorithms with no effort
  • Automatic differentiation: how to let the computer code your algorithm 
  • Application to quantum Monte Carlo : wavefunction value and local energy differentiation
  • The course will be supplemented by class exercizes of general interest  (e.g. adjoint of matrix-matrix, Choleski decomposition, diagonalization)