Table 5 Single-layer training performance (in terms of epoch duration) for different publicly available forward-only and local learning repositories, compared to SCFF, on CIFAR-10

From: Self-Contrastive Forward-Forward algorithm

Method

Unsupervised

Epoch duration (s)

Acc. (%)

SigProp (Kohan et al. 202346)

✗

7.5

91.6

HardHebb (Miconi et al. 202153)

✓

1.6

64.8

HardHebb (Lagani et al. 202279)

✓

5.1

64.6

SoftHebb (Journé et al. 202216)

✓

2.6

80.3

SCFF (ours)

✓

1.2

80.8

  1. Accuracy results of the overall models from the respective papers on CIFAR-10 are also reported. All experiments were conducted on a single NVIDIA RTX 4090 GPU.
  2. Entries in bold indicate the proposed method and its associated results.
Back to article page