ComplexNet is an advanced and flexible artificial neural network framework designed to handle complex-valued parameters. Unlike traditional real-valued neural networks, CNet explores the idea of utilizing complex numbers in the field of machine learning. This library is currently a work in progress, with ongoing development to harness the efficiency
The implementation integrates the Strassen algorithm for matrix multiplication, reducing computational complexity and enhancing the matrix multiplication. SIMD (Single Instruction, Multiple Data) instructions accelerate general matrix operations, improving overall computational throughput. OpenMP Optimization (OMP) parallelizes various tasks in the framework, effectively distributing the workload across multiple cores. These features collectively contribute to delivering a good-performance neural network framework, well-suited for demanding machine learning tasks.
- Devansh. (2022, November 17). Improve neural networks by using complex numbers - geek culture - medium. Geek Culture. https://medium.com/geekculture/improve-neural-networks-by-using-complex-numbers-5e142b8931e6
- Preprint, A., Ko, M., Panchal, U. K., Andrade-Loarca, H., & Mendez-Vazquez, A. (n.d.). Coshnet: A hybrid complex valued neural network using shearlets. Arxiv.org. Retrieved January 17, 2024, from http://arxiv.org/abs/2208.06882
- Panchal, U. (n.d.). Coshnet. https://github.com/Ujjawal-K-Panchal/coshnet