Scalable simulation of open quantum systems requires efficient treatment of non-Markovian environments. Dissipation-Assisted Matrix Product Factorization (DAMPFf) combines tensor networks with pseudomode mapping to recast the integrodifferential dynamics into a time-local Lindblad form. At each time step, dampf performs O(N 5) singular-value decompositions (SVDs) to compress the MPS—this dominates the runtime. In this thesis, I replace deterministic SVD with a randomized SVD (RSVD) scheme and demonstrate that RSVD (with minimal oversampling) retains relative errors below 10−4 while cutting the cost of each factorization by up to 50%. Across benchmark problems, this yields overall simulation speedups of up to 2× without compromising accuracy. My results extend dampf’s reach to larger systems and longer timescales in open-system dynamics.     «

Scalable simulation of open quantum systems requires efficient treatment of non-Markovian environments. Dissipation-Assisted Matrix Product Factorization (DAMPFf) combines tensor networks with pseudomode mapping to recast the integrodifferential dynamics into a time-local Lindblad form. At each time step, dampf performs O(N 5) singular-value decompositions (SVDs) to compress the MPS—this dominates the runtime. In this thesis, I replace deterministic SVD with a randomized SVD (RSVD) scheme and de...     »