When do we use a high value of momentum in Neural Network Forecasting? To avoid overfitting the dataset. To fit local patterns of the data. To increase the number of epochs. To avoid local minimum. To reduce the learning rate of the algorithm