Table 13 Computation-time summary.

From: Kolmogorov–Arnold Networks for predicting carotid intima-media thickness in cardiovascular risk assessment

Model

Training time (GPU h)

Inference latency for GPU (ms/record)

Inference latency for CPU (ms/record)

Kolmogorov–Arnold Network (KAN)

0.62 ± 0.03

1.8 ± 0.2

14.5 ± 1.1

Deep Neural Network

0.28 ± 0.02

1.5 ± 0.1

12.7 ± 0.9

Multi-layer perceptron

0.12 ± 0.01

1.3 ± 0.1

11.9 ± 1.0

Random Forest (CPU only)

–

–

9.6 ± 0.8

Back to article page