Modal Interval Arithmetic
ALiX™ is the only Artificial Intelligence (AI) engine based on our patented Modal Interval Arithmetic (MIA). All other AI engines are based on Floating Point Arithmetic (FPA). So, why is this an important differentiator?
Accurate and Repeatable Results
FPA is a method of using finite approximations of real numbers to perform calculations in a computer. This approach has accuracy and rounding error issues that manifest themselves in the form of uncertain and non-repeatable outcomes.
MIA does not have rounding error issues and produces reliable and repeatable outcomes.
Advanced numerical methods based on FPA, such as stochastic gradient descent, are primarily a one-step-at-a-time, or “sequential,” process. This means they cannot take advantage of parallelism, which is the ability to scale to a large number of machines when computing an outcome. This is a fundamental issue, because AI typically requires processing very large databases to obtain desired results.
ALiX, based on MIA, is highly parallel, meaning the processing of large volumes of data can be easily distributed over a large number of machines for improved speed and performance.
Reliable Non-Linear Optimization
Most AI processing requires a calculation called non-linear optimization. Methods based on FPA cannot reliably find solutions to non-linear optimization problems because they are heuristic and/or statistical in nature. Such methods are, in effect, only a “best guess.”
MIA reliably solves non-linear optimization calculations in a deterministic manner, eliminating the guesswork and providing confidence and certainty in outcomes.