Predictive coding is a famous and influential theory in the field of neuroscience. It states that instead of passively receiving the external information and forming perception or decision, the brain holds a hierarchical internal modal that could actively interact with the external stimuli. In this hierarchy, each level could predict the activation of the lower level, with the lowest level representing the outside world; while the predicted level could compute the difference between real stimulation and the prediction and send it upward to update the model for better prediction in the future. It could provide explanations for a wide range of neurophysiological and psychological phenomena, which leads to the belief that predictive coding may serve as a unifying computational framework for brain functions including sensation, perception, memory, and so on. However, the interpretation provided by predictive coding on brain function is not definite enough. More supportive evidence is needed to establish predictive coding as a unifying principle used by the brain. On one hand, researchers build computational models implementing predictive coding dynamics and check whether such a dynamic system could generate some similar results observed in the biological brain. On the other hand, efforts have been made to investigate the neural mechanisms of predictive processing in the brain. In the current thesis, we combined both sides, experimental and computational, attempting to provide more solid evidence for predictive coding as a unifying theory of brain function. Specifically, we conducted three progressive and related studies. The first one focuses on the neural activities in the biological brain, in particular, the oscillatory traveling waves. Could cortical traveling waves underlie predictive processes in the brain? Our results suggest that the ascending traveling waves only appear with the presence of bottom-up driven visual stimuli and disappear when visual inputs are absent; the descending waves, although they receive some modulation from external visual input, are less affected. We explained the results in the predictive coding framework: oscillatory traveling waves might be a neural signature of predictive coding with forward waves carrying prediction errors and backward waves transmitting prediction signals. In the second project, we utilized the deep learning technique and constructed a neural network which could implement predictive coding dynamics. If the human brain functions according to such a dynamic principle, the predictive coding network should show certain human-like properties. We tested this hypothesis with illusory contour image stimuli. The study suggests the neural network driven by predictive coding dynamics possesses illusory perception, supporting the possibility that the same dynamics strategy, i.e. predictive coding, might be shared between the network and biological brain. Based on the results of the first two studies: (i) cortical traveling waves reflecting the neural mechanisms of predictive coding in the biological brain; (ii) the predictive neural network displaying human-like performance, in the third study, we further update the same predictive neural network by adding biologically plausible time delays and constants between network layers in order to generate oscillations.[...]