Successfully and efficiently training deep multi-layer perceptrons with logistic activation function simply requires initializing the weights with an appropriate negative mean
Ahmet Yılmaz, Riccardo Poli
Topics & Concepts
Sigmoid functionInitializationActivation functionArtificial neural networkLogistic functionPerceptronComputer scienceBackpropagationGradient descentFunction (biology)Artificial intelligenceObstacleTraining (meteorology)AlgorithmLayer (electronics)Multilayer perceptronMathematicsApplied mathematicsMachine learningPhysicsLawChemistryOrganic chemistryEvolutionary biologyProgramming languageMeteorologyBiologyPolitical scienceNeural Networks and ApplicationsMachine Learning and ELMDomain Adaptation and Few-Shot Learning