The Neural Control of the Kidneys

Nerve regulation of the kidney is mediated through the sympathetic nervous system which identified distended renal sympathetic innervation and neuro-effector binding along granulosa cells, renal tubules, and adrenal glands of the renal vasculature (Kate et al., 2004; Edward Et al., 2004; J., 2011). The role of the sympathetic nervous system is to release norepinephrine into the interstitial space at the sympathetic end. Since moderately high levels of neural stimulation have a greater effect on centrifugal arterioles than afferent nerve stimulation, RBF declines more than GFR, consistent with efferent arteriolar contraction, but afferent vasoconstriction It dominates during maximum neural stimulation.

Identification Neural Computer (DNC) is a neural network that learns from the beginning to form and use complex memory structures such as lists, trees, graphics and so on. The core of the neural network is called a controller, and behaves like a processor in a conventional computer by receiving inputs from the memory, reading and writing, and generating output. When training a DNC, the controller learns to produce better answers by using the correct memory structure. Genetic programming and evolution strategy (ES) is an optimization method decades ago that has been used to simulate natural evolution without using back propagation. After all, evolution is the only known strategy that at least produces human intelligence. Evolutionary strategies provide a range of candidate solutions for a given problem and evaluate them based on the "fitness" function.

About 28 years ago, a partnership between DARPA and Carnegie Mellon University developed a neural network based vehicle control system called ALVINN. It is a compact neural network that uses parameters less than 50 kB and achieves 90% accuracy only under simulation conditions. However, this research is novelty of neural networks that have not been explored for nearly 30 years. End-to-end paradigm Of course, this novelty is end-to-end automatic operation. The end-to-end autopilot is a design paradigm that directly calculates the steering and acceleration of raw or minimal processed sensor data, including video and LIDAR information. In this example we do not use the AI ​​subsystem (such as the lane detector) to notify the deterministic navigator, but we allow the vehicle to learn the whole autopilot task from the beginning. To understand how it works, we first need to check where the existing system failed.