optimization
9 posts tagged optimization.
- Optimizers Are Becoming Systems Choices, Not AdamW Replacements
- A Newton Method That Hits O(1/k³) With One Linear Solve Per Step
- Branch-and-bound gets practical for messy robot inspection routes
- AlphaEvolve nudges the matrix multiplication exponent lower
- LP-NAS puts a linear program inside differentiable architecture search
- Muon groks faster, then the readout drifts
- Hyperball optimizers still need learning-rate discipline
- Barzilai-Borwein’s superlinear convergence problem
- Adam Was the Default, Not the Answer, for MLIP Training