Resampling Methods in Neural Networks: From Point to Interval Application to Coordinate Transformation

In the development of neural networks, many realizations are performed to decide which solution provides smallest prediction error. Due inevitable random errors associated with data and randomness related network (e.g., initialization weight initial conditions linked learning procedure), there is usually not an optimal solution. However, we can advantage idea making several based on resampling methods. Resampling methods often used replace theoretical assumptions by repeatedly original inferences from resampling. provide us opportunity do interval instead only one point prediction. Following this idea, introduce three in namely Delete-d Jackknife Trials, Delete-1 Hold-Out Trials. They discussed applied a real coordinate transformation problem. Although Trials offer better results, choice method will depend dimension problem at hand.