notesum.ai

Published at November 27

Preserving Information: How does Topological Data Analysis improve Neural Network performance?

cs.NE

Released Date: November 27, 2024

Authors: A. Stolarek, W. Jaworek

Arxiv: http://arxiv.org/abs/2411.18410v1

Refer to caption

Model	Training Set Size	Accuracy	Confidence Interval
TDA	100	0.1880 $\pm$ 0.0340	(0.1669, 0.2091)
TDA	250	0.2500 $\pm$ 0.0456	(0.2217, 0.2783)
TDA	500	0.3060 $\pm$ 0.0803	(0.2562, 0.3558)
TDA	1000	0.3420 $\pm$ 0.0540	(0.3085, 0.3755)
RAW	100	0.2600 $\pm$ 0.0874	(0.2058, 0.3142)
RAW	250	0.2370 $\pm$ 0.0936	(0.1790, 0.2950)
RAW	500	0.1860 $\pm$ 0.0602	(0.1487, 0.2233)
RAW	1000	0.1670 $\pm$ 0.0452	(0.1390, 0.1950)
V-S	100	0.6930 $\pm$ 0.0683	(0.6507, 0.7353)
V-S	250	0.7740 $\pm$ 0.0676	(0.7321, 0.8159)
V-S	500	0.8110 $\pm$ 0.0643	(0.7712, 0.8508)
V-S	1000	0.8220 $\pm$ 0.0414	(0.7963, 0.8477)