Statistical Mechanics of On-line Ensemble Teacher Learning through a Novel Perceptron Learning Rule

摘要: In ensemble teacher learning, teachers have only uncertain information about the true teacher, and this is given by an consisting of infinite number whose variety sufficiently rich. a student learns from that iteratively selected randomly pool many teachers. An interesting point learning asymptotic behavior to approach The performance improved using Hebbian rule in learning. However, perceptron cannot improve performance. On other hand, we proposed with margin. This identical when margin zero infinity. Thus, connects cont...