@misc{9244213,
  abstract     = {{Modern Neural Processing Units (NPUs) must be evaluated long before silicon ex-
ists, yet early tools force a difficult compromise: fast analytical performance models
enable exploration, but may diverge from slower RTL simulation, which provides
a more accurate timing reference. This thesis investigates whether that gap can be
learned and corrected. Using early-available block-command descriptors, sequence
summaries, and analytical cycle estimates from Arm’s NPU modelling flow, super-
vised XGBoost regressors are trained as a correction layer rather than a replace-
ment for the existing model. The method is evaluated on anonymised single-block
and short-sequence workloads across three NPU units, comparing corrected pre-
dictions against RTL cycle measurements with a 5% acceptance criterion. Results
show large improvements in difficult cases, raising several low baseline pass rates
above 90% while substantially reducing median and tail errors. SHAP-based ana-
lysis further identifies workload regimes linked to systematic mismatch, making
the correction model not only more accurate but also a practical diagnostic tool
for model improvement.}},
  author       = {{Gouveia, Joaquim and Johansson, Oliver}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Machine-Learning-Based Correction of Analytical NPU Performance Models}},
  year         = {{2026}},
}