On the convergence of adaptive first order methods: proximal gradient and alternating minimization algorithms