Methods for learning the parameters of a convolutional neural network, and for classifying an input datum

A method for classifying an input datum, comprising the implementation of steps of:

• (a) Learning, by a data processor of at least one server, the parameters of a third CNN comprising a last fully-connected layer, comprising steps of:
• (a1) Learning, from a base of already-classified confidential learning data, the parameters of a first CNN;
• (a2) Learning, from a base of already classified public learning data, the parameters of a last fully-connected layer of a second CNN corresponding to the first CNN to which said fully-connected layer has been added;
• (a3) classifying data of a second base of as yet unclassified public data, by means of the second CNN;
• (a4) learning, from the second base of public learning data now classified, the parameters of the third CNN;
• (b) Classifying said input datum, by a data processor of a terminal, using the third CNN.

Non-transitory computer readable medium storing a program causing a computer to execute a method for classifying an input datum, the method comprising the implementation of steps of:

• (a) Learning the parameters of a third CNN comprising a last fully-connected layer, comprising:
• (a1) Learning, from a base of already-classified confidential learning data, the parameters of a first CNN;
• (a2) Learning, from a base of already classified public learning data, the parameters of a last fully-connected layer of a second CNN corresponding to the first CNN to which said fully-connected layer has been added;
• (a3) classifying data of a second base of as yet unclassified public data, by means of the second CNN;
• (a4) learning, from the second base of public learning data now classified, the parameters of the third CNN;
• (b) Classifying said input datum using the third CNN.