Skip to main content
AI Perspectives & Advances

Table 6 The grid search range of hyperparameters for experiment 2

From: CLeaR: An adaptive continual learning framework for regression tasks

Hyperparameter

Value

Dropout

{0,0.005}

Novelty buffer size

{512,1024}

Latent size

{4,5}

Threshold factor

{0.75}

Online EWC γ

{0.7,0.8,0.9}

Online EWC Îğa

{200}

Online EWC Îğp

{600,800}

  1. The parameter Îğa is the lambda value for the autoencoder, and Îğp is for the predictor. Latent size is the dimension of the bottleneck in the autoencoder
Back to article page