This is an interesting way to think about how to get to a minimal form of a complex system.
A friend in college told me of a research project that had managed to balance a simulated inverted pendulum in 2D using 25 neurons and back propagation. But I had done this exact problem with conventional state space controls using only 5 summations (the equivalent of 5 neurons).
After I finish patting myself on the back, you then wonder what it would take for that 25 neuron solution to keep optimizing down the theoretical 5 neuron solution? The article is an interesting approach to that problem.
The paper they reference used 3456 input neurons and 9 output neurons, with no hidden nodes. They designed their input and output differently, so it's not a direct comparison. The optimized solution has 17 inputs, 2 outputs, and 2 hidden nodes. That's a massive level of optimization.
This sounds like the feed of a single male. Facebook showing sleazy content/ads to single guys predates AI by a lot. Try removing your single relationship status from your profile and see what changes.
Whenever you drive/walk in soft terrain, the wheel/leg is constantly climbing the ramp created by it sinking into the terrain. In a perfect system, this determines how much power you need to move. This is why trains are so efficient. A hard wheel on a hard rail has very little deflection -- so excellent efficiency.
Wheels have to climb that ramp for every inch of travel. Legs get to step forward and only take that penalty for each step. If everything else is the same, the legs win on soft terrain.
But everything else is never the same :-). The early legged vehicles used linear motions, which means you have these very long sliding surfaces. This is heavy and the drive system efficiency dominates over the terrain interaction efficiency. Add in the fact that you have multiple axis to drive and the weight and drive losses really add up.
Modern dog and human style walking robots are MUCH better on efficiency than those early designs. However, they require enough sensing and compute to dynamically balance. Legs can do things that wheels can't, but you have to have smart enough software to take advantage of that. The compute available for a high radiation environment is a fraction of what is in your smartwatch. Wheels are still winning on energy efficiency, but at least it's getting closer.
I worked on Dante at CMU and Marsokhod at NASA Ames; and was in the same group that developed Ambler.
Watch out of patent problems. There was a major dust up between Sonos and Google over audio sync technology.
Disclaimer: I've worked for both companies, but not on that
This talks about the "what" of the code, but you have to also convey the "why".
If you have a well understood problem space and a team that is up to speed on it, then the "why" is well established and the code is the "what".
However, there are often cases where the code is capturing a new area that isn't fully understood. You need to interleave an education of the "why" of the code.
I was once asked to clean up for release 10k lines of someone else's robotics kinematics library code. There weren't any comments, readmes, reference programs, or tests. It was just impenetrable, with no starting point, no way to test correctness, and no definition of terms. I talked to the programmer and he was completely proud of what he had done. The variable names were supposed to tell the story. To me it was a 10k piece puzzle with no picture! I punted that project and never worked with that programmer again.
That depends on if it's in the grass that the cows eat. They first researched iodine in cows because there isn't much iodine in the grass around the great lakes. This area was known as the "goiter belt".
Adding iodine to the cow's food made them healthier and that's how they estimated the dose for humans (mcg / kg).
A friend in college told me of a research project that had managed to balance a simulated inverted pendulum in 2D using 25 neurons and back propagation. But I had done this exact problem with conventional state space controls using only 5 summations (the equivalent of 5 neurons).
After I finish patting myself on the back, you then wonder what it would take for that 25 neuron solution to keep optimizing down the theoretical 5 neuron solution? The article is an interesting approach to that problem.
The paper they reference used 3456 input neurons and 9 output neurons, with no hidden nodes. They designed their input and output differently, so it's not a direct comparison. The optimized solution has 17 inputs, 2 outputs, and 2 hidden nodes. That's a massive level of optimization.