• In the field of computer vision, common feature amount includes SIFT, SURF, HOG, etc., using these feature amounts to the image data into a vector, then use the machine learning classifier vector after conversion.
• Loss of function is a function parameter derivation of the loss function, expressed "if slightly change the value of this parameter, the value of the loss function of how change."
If the value of the derivative is negative, by changing the parameter toward a positive direction, decrease the value of the loss function,
If the value of the derivative is positive toward the negative direction by changing the parameter, the value of the loss of function can be reduced.