At the recent Genetic and Evolutionary Computation Conference (GECCO 2024) in Melbourne, Australia, Associate Professor Ran CHENG’s research team from the Department of Computer Science and Engineering (CSE) at the Southern University of Science and Technology (SUSTech) was awarded the Best Paper Prize.

Their paper was titled “Tensorized NeuroEvolution of Augmenting Topologies for GPU Acceleration”, with master’s student Lishuang WANG of the Department of CSE serving as the first author and Associate Professor Ran CHENG as the corresponding author.

The NeuroEvolution of Augmenting Topologies Algorithm (NEAT) has significantly impacted fields such as artificial intelligence, robot control, and autonomous driving. However, traditional NEAT algorithms face computational efficiency challenges when dealing with large-scale tasks.

To address these limitations, Ran CHENG’s team developed the TensorNEAT algorithm library. Utilizing tensor quantization technology, NEAT and its derivative algorithms (including CPPN and HyperNEAT) can fully support GPU acceleration.

Tensorization, a technique that converts data structures and operators into tensor form, is particularly well-suited for efficient parallel computing on GPUs. TensorNEAT transforms the adaptable network topology in the NEAT algorithm into tensor form, allowing key operations to be executed in parallel across the entire population, thus significantly enhancing the algorithm’s computational efficiency.

Experimental results demonstrate that TensorNEAT is up to 500 times faster than traditional NEAT on various tasks and hardware configurations.

Since its inception in 1999, GECCO has become the flagship conference in evolutionary computing, renowned for its academic quality and influence. Hosted by ACM SIGEVO, the conference annually attracts the world’s leading researchers and scholars to exchange and present the latest advancements in evolutionary computing.