• webmaster9757

Warp Speed Ahead

by Justin Lapre


In the first few pages of "The Physics of Star Trek" by Lawrence M. Krauss (yes, I actually do own a copy) the crew is killed almost immediately. The culprit turns out to be nothing more sinister than our old friend inertia. You feel it when you accelerate from 0 to 60 mph in 3.5 seconds while driving your Tesla Roadster: you get pushed back into your seat because your body wants to stay put. So the writers created the idea of inertial dampeners allowing ships to accelerate to light speed and back at a moment's notice. Accelerating to the speed of light would take months if you were taking precautions to not liquefy your crew members.

Months just happens to be the amount of time businesses don't have to spare to make business-critical decisions. To get a bit technical, some fairly common and interesting problems fall into the non-deterministic (NP) category such as factoring primes which is useful for encryption. NP simply means that potential solutions can be verified quickly; finding such a solution can take a really long time. For example, finding the prime factors for 10,967,535,067 could take a while. But if I tell you that 104,723 and 104,729 are the factors, one can easily verify the truth of this claim.

Factoring is a fairly binary problem: numbers either are or are not factors. Other problems, such as the traveling salesman problem (TSP), may better lend themselves to approximation. TSP is a problem of minimizing the path taken for a salesman to visit n different cities exactly once (presumably to limit the outrageous cost of air travel). It's probably better to get a "pretty good" solution in a minute than the perfect solution in 1,000 years.

When you run your machine learning models on the Lucd platform, you don't need to worry about the details of ensuring everything is done in a fast and efficient manner (while still maintaining correctness). It was built from the ground-up to to take care of all of that for you by constantly maximizing the available parallelism inherent in your model. We want to help guide your model as close to light speed as possible. Lucd has solved these problems; now you are free to just have fun with your model.