latifaboubekri-hub/topsis-gpu-acceleration: Initial release

First public release of GPU-TOPSIS, a fully vectorized GPU-accelerated reformulation of the TOPSIS method for large-scale Multi-Criteria Decision Making. This release includes the complete experimental notebook used in the paper, covering four TOPSIS implementations (CPU/NumPy, CuPy, PyTorch, TensorFlow) and a two-pass sharding algorithm enabling the processing of decision matrices up to 200 million alternatives on a single GPU. All experiments are reproducible on Google Colab Pro (Tesla T4) using the Amazon Products 2023 dataset. Highlights:

Up to 9.43× speedup over CPU baseline at million-scale Memory-safe sharding with mathematically exact rankings (Property 1) 9 experiments covering scalability, numerical consistency, sensitivity analysis, and ultra-large-scale evaluation (100M–200M alternatives) Validated across three independent GPU backends (CuPy, PyTorch, TensorFlow)