High Performance Computing
Papers:
- Neural Acceleration of Graph Based Utility Functions for Sparse Matrices [file:][pdf:]
- Neural Acceleration for Incomplete Factorization Preconditioning [file:][pdf:]
- Grokking - Generalization Beyond Overfitting on Small Algorithmic Datasets - 2201.02177v1 [file:][pdf:]
- Grokfast - Accelerated Grokking by Ampilfying Slow Gradients - 2405.20233v2 [file:][pdf:]
- Evaluation of Pre-training LLMs on Leadership Class Supercomputers [file:][pdf:]
- papers/Probability-Aware Selective Protection for Sparse Iterative Solver.pdf
- papers/Fiedler's Theory of Spectral Graph Partitioning.pdf
- books/Spectral Algorithms.pdf
To read:
- Symbolic Factorization of Sparse Matrix
- RCM
- Approximate Minimum Degree Ordering Algorithm (AMD)
- Nested Dissection (ND)
- Graph Based Utility Functions
- vertex information
- graph partition
- subgraph identification
- Spectral Subarea convolution operator
- Spatial Subarea convolution operator
- ILU(k)
- ICHOL(k)
- Conjugate Gradient Method
- Generalized Minimal Residual Method
- Transformer Architecture [pg. 3]