Just imagine your root function as: F(f_1(x_1), f_2(x_2), …, f_n(x_n)) The root function is a collection of projections acting on (x_1, x_2, …). And the comp

Just imagine your root function as: F(f_1(x_1), f_2(x_2), …, f_n(x_n)) The root function is a collection of projections acting on (x_1, x_2, …). And the components receive the right set of coordinates to act on. I’m not sure if it’s clearer that way though.