So are you saying ReLu based nets are optimizable by removing all need for Relu activation functions?
Are you saying in your other post we can optimize each layer? An input has an output in a trained net. If we randomly sample many times, maybe we can build a heterarchy and throw the net in the trash? It would let us see the relationship between various input, outputs, and inputs to outputs. Better optmizations could lower cost or keep the same cost and speed while being less RAM hungry. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T894f73971549b2ee-M6a1948e1ff99a817b659930a Delivery options: https://agi.topicbox.com/groups/agi/subscription
