DeepMind's AlphaDev: Uncovering Faster Sorting Algorithms through DRL

DeepMind's AlphaDev has used deep reinforcement learning (DRL) to discover new and more efficient fixed and variable sorting algorithms than the state-of-the-art human benchmarks. The algorithms were discovered by playing a single-player game called AssemblyGame, in which the player selects a series of low-level CPU instructions to combine to yield a new and efficient sorting algorithm. The discovered fixed and variable sort algorithms were found to be both new and more efficient than the state-of-the-art human benchmarks. The research also tested the generality of AlphaDev and found that it can optimize non-trivial, real-world algorithms. Additionally, the discovered algorithms have been incorporated into the LLVM C++ library, used by millions of developers and applications around the world. The discovered algorithms have led to improvements of up to 70% for sequences of a length of five and roughly 1.7% for sequences exceeding 250,000 elements for the uint32, uint64 and float data types for ARMv8, Intel Skylake and AMD Zen 2 CPU architectures.

#AI #OpenAI