Table 4 Test accuracy [%] comparison between SCFF and energy-based methods with local learning rules on the MNIST, CIFAR-10, and STL-10 datasets

From: Self-Contrastive Forward-Forward algorithm

Method

Unsupervised

MNIST

CIFAR-10

STL-10

EqProp (Laborieux et al. 202124)

✗

98.0

88.6

–

EqProp (Liu et al. 202482)

✓

97.6

71.5

–

DualProp (Høier et al. 202322)

✗

98.4

92.3

–

SCFF (ours)

✓

98.7

80.8

77.3

  1. The symbols ✓, ✗, and - mean “yes", “no", and “no reported results", respectively.
  2. Entries in bold indicate the proposed method and its associated results.
Back to article page