resLIK (version 0.1.2)

agreement: Agreement Sensor

Description

The Agreement Sensor measures the alignment between two latent representations, typically from different modalities or redundant sensors. It uses cosine similarity to quantify agreement.

Usage

agreement(z1, z2, eps = 1e-08)

Value

A list containing:

agreement

The cosine similarity score (-1 to 1).

Arguments

z1

Numeric vector or matrix. The first representation.

z2

Numeric vector or matrix. The second representation. Must have the same shape as z1.

eps

Numeric. Small constant for numerical stability. Defaults to 1e-8.

Details

Agreement Sensor

The sensor computes the cosine similarity between z1 and z2: $$A = \frac{z_1 \cdot z_2}{||z_1|| ||z_2|| + \epsilon}$$

Values close to 1 indicate strong agreement, 0 indicates orthogonality (no agreement), and -1 indicates opposition. In the context of RLCS, high positive agreement is generally required for a PROCEED signal.

Examples

# Example 1: Perfect Agreement
z1 <- c(1, 2, 3)
z2 <- c(2, 4, 6)
agreement(z1, z2)

# Example 2: Disagreement (Orthogonal)
z3 <- c(1, 0, 0)
z4 <- c(0, 1, 0)
agreement(z3, z4)