Today was the day when a Neural-Network based learning controller could successfully accomplish a complicated task for which very little information was known analytically. This potentially has phenomenal implications in deriving us, the engineers, to make systems intelligent to learn on their own. We can more confidently leave them unsupervised. To my best knowledge, no one else before had used NN in a real time control loop for a complicated task. "Complicated" in a sense that no mathematical modeling of the system was ever introduced. I believe it is not just an Adaptive Controller which learns to adopt its settings only to compensate for graduate changes in the input conditions. It is one degree smarter: It is a kind of adaptive control which can improve its initial performance over time, even if there is no change in its environment. In other words, it is enough only to start with settings based on which the system doesnt go unstable (That little initial info are certainly always needed). Then mere running will help it learn after a couple of runs.
Anyways, I was excited really today. When thought why should I be? I realized that it is good to be excited about your work. That is how this world likes. When I was younger and had an illusion of the "truth' to be something extraordinary and big, no such an achievements were making me excited. I kept them inside, thinking these are nothing compared to real truth. Now, thu, I rather show this to the emotional world so, it can help me take bigger strides next. So, here I am, full of hopes, no fear of loss, and confident of the path I am walking along.
congratulation my blood brother !!!
ReplyDeletekeep moving forward