Does anyone here have a simple vanilla RNN that uses Backprop that I can compare to my simple Markov Chain? The RNN must not have any semantics, residuals, gates, ensemble, data augmentation, etc, _*only *_the ability to predict text (must have order I suppose, hence RNN) and the ability to use Backprop to get any clue how to predict. I'm trying to investigate if (or how much better) is Backprop than a no backprop net. And why (what rules Backprop learns).
Another question: Is backprop learning rules like XOR? What about physics simulations like seeing reflections off wood that in actuality show a cat face that isn't visible to the naked eye? (there's such an algorithm in its early stages) Or will Backprop in a vanilla style only going to learn syntax/ backoff/ random forests/ dropout? ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2bdb1204cee002e0-Mba28cbe5c58909f217380dc2 Delivery options: https://agi.topicbox.com/groups/agi/subscription
