ReOS-ELM and FOS-ELM

ReOS-ELM and FOS-ELM are not separate classes in v2. They are toggles on RlsOptions<FloatT>.

ReOS-ELM

ReOS-ELM regularizes the initial covariance used by the online solver.

feature_elm::RlsOptions<float> options;
options.regularization = 1e-2f;
options.forgettingFactor = 1.0f;
options.constraint = feature_elm::RlsConstraint::kNone;

Use it when:

• The initial chunk is small.
• The hidden feature matrix is ill-conditioned.
• Early online updates produce unstable weights.

The extension is off when regularization = 0.

FOS-ELM

FOS-ELM applies a forgetting factor to down-weight stale data.

feature_elm::RlsOptions<float> options;
options.regularization = 0.0f;
options.forgettingFactor = 0.98f;
options.constraint = feature_elm::RlsConstraint::kNone;

Use it when:

• The stream distribution changes over time.
• Recent labels are more informative than old labels.
• You need to compare against ordinary OS-ELM.

The extension is off when forgettingFactor = 1.

Drift comparison

FOS-ELM should recover faster after the configured drift point because older chunks receive lower weight.

References

• Liang, Huang, Saratchandran, and Sundararajan. 2006. OS-ELM.
• Online sequential learning and forgetting-factor variants in the ELM literature.