In this paper, we propose a non-parametric probabilistic load flow (NP-PLF) technique based on Gaussian process (GP) learning. The technique can provide "semi-explicit" power flow solutions by implementing learning step and testing step. The proposed NP-PLF leverages upon GP upper confidence bound (GP-UCB) sampling algorithm. The salient features of this NP-PLF method are: i) applicable for power flow problem having power injection uncertainty with unknown class of distribution; ii) providing probabilistic learning bound (PLB) provides control over the error and convergence; iii) capable of handling intermittent distributed generation as well as load uncertainties; and iv) applicable to both balanced and unbalanced power flow with different type and size of systems. The simulation results performed on IEEE 30-bus and IEEE 118-bus system show that the proposed method is able to learn the state variable function in the input subspace using a small number of training samples. Further, the testing with different distributions indicates that more complete statistical information can be obtained on probabilistic power flow problem using the proposed method.