METHODS FOR LEARNING AND TESTING A MULTI-LAYER PERCEPTRON BASED MARINE ENVIRONMENT INFORMATION CORRECTION MODEL FOR CORRECTING MARINE ENVIRONMENT INFORMATION USING WAVE PREDICTION INFORMATION AND DEVICES USING THE SAME

摘要

In a method of learning a multi-layer perceptron-based marine environment information correction model that corrects marine environment information using wave prediction information, (a) a learning device includes: a first past time point to m-th where m is 2 or more Integer - at each of the past time points, the first wave prediction information to the mth wave prediction information for predicting the wave of the specific area using each of the first atmospheric environment information to the mth atmospheric environment information of the specific area, and the mth wave prediction information in the specific area 1st to nth - where n is an integer greater than or equal to 2 - With reference to 1_1st to m_nth observation information that observed marine environment information at each point, i - the i is an integer greater than or equal to 1 and less than or equal to m - past j-th point of time - wherein j is an integer greater than or equal to 1 and less than or equal to n - obtaining, as a training data set, i_j-th wave prediction information, which is wave prediction information at the point, and i_j-th observation information, which is observation information corresponding thereto; (b) the learning apparatus inputs the i_j wave prediction information into a marine environment information correction model, and causes the marine environment information correction model to perform a neural network operation on the i_j wave prediction information using a preset parameter to calculate the ith outputting the i_j th marine environment correction information that predicts the correction result of the marine environment correction information at the j th point in the past; and (c) the learning device generates a loss by referring to the i_jth marine environment correction information and the i_jth observation information, and updates the parameter of the marine environment information correction model through backpropagation using the loss learning the marine environment information correction model by doing so; Disclosed are a method and a learning apparatus comprising: