torch.distributions.transforms.AbsTransform

class AbsTransform extends Transform

new AbsTransform(options?: TransformOptions)

Absolute value transform: maps ℝ → [0, ∞).

Transform via the mapping y = |x|.

Note: This transform is NOT bijective since it maps both positive and negative values to the same output. The inverse returns the positive root.

\begin{aligned} \text{Forward: } y = |x| \\ \text{Inverse: } x = y \quad \text{(returns positive root)} \\ \text{Log-det Jacobian: } \log\left|\frac{dy}{dx}\right| = 0 \\ \text{Domain: } \mathbb{R}, \text{Codomain: } [0, \infty) \end{aligned}

torch.distributions.transforms.AbsTransform

readonlydomain(unknown)

readonlycodomain(unknown)

readonlybijective(unknown)

readonlysign(unknown)

Absolute value transform: maps ℝ → [0, ∞).

Transform via the mapping y = |x|.

Note: This transform is NOT bijective since it maps both positive and negative values to the same output. The inverse returns the positive root.

\begin{aligned} \text{Forward: } y = |x| \\ \text{Inverse: } x = y \quad \text{(returns positive root)} \\ \text{Log-det Jacobian: } \log\left|\frac{dy}{dx}\right| = 0 \\ \text{Domain: } \mathbb{R}, \text{Codomain: } [0, \infty) \end{aligned}