This work proposes a tree tensor network without activation nonlinearity, and for the first time reveals its equivalence to quantum-inspired models.