Large N reduced models and stochastic quantization