Table 3 Comparison on model training time of CNNs trained with different image representations.

From: Converting tabular data into images for deep learning with convolutional neural networks

Data

Comparison

Log2 ratio of model training time

P-value

CTRP

REFINED vs. IGTD

− 0.034 (0.232)

5.35E−01

DeepInsight vs. IGTD

4.136 (0.240)

5.55E−25

GDSC

REFINED vs. IGTD

0.172 (0.199)

1.30E−03

DeepInsight vs. IGTD

4.622 (0.417)

2.34E−21

  1. The number before parenthesis is the average across 20 cross-validation trials, and the number in the parenthesis is the standard deviation. P-value is obtained via the two-tail one-sample t-test across the cross-validation trials.
Back to article page