Lines Matching refs:scale
37 `scale` matrix: `covariance = scale @ scale.T`, where `@` denotes
43 `scale` matrix + `loc`, applied to the positive half-space:
44 `Supp = {loc + scale @ x : x in R^k, x_1 > 0, ..., x_k > 0}`. On this set,
47 pdf(y; loc, scale) = exp(-||x||_1) / Z, for y in Supp
48 x = inv(scale) @ (y - loc),
49 Z = |det(scale)|,
55 * `scale` is a linear operator in `R^{k x k}`, `cov = scale @ scale.T`,
59 The VectorExponential distribution is a member of the [location-scale
65 Y = (Y_1, ...,Y_k) = scale @ X + loc
79 mean and covariance (by setting `loc` and `scale`), while preserving some
128 The `batch_shape` is the broadcast shape between `loc` and `scale`
132 `scale`. The last dimension of `loc` (if provided) must broadcast with this.
134 Recall that `covariance = scale @ scale.T`.
137 scale = diag(scale_diag + scale_identity_multiplier * ones(k))
148 `scale` is the Identity matrix.
155 matrix added to `scale`. May have shape `[B1, ..., Bb, k]`, `b >= 0`,
157 `scale`. When both `scale_identity_multiplier` and `scale_diag` are
158 `None` then `scale` is the `Identity`.
160 a scaled-identity-matrix added to `scale`. May have shape
162 `k x k` identity matrices added to `scale`. When both
163 `scale_identity_multiplier` and `scale_diag` are `None` then `scale` is
185 scale = distribution_util.make_diag_scale(
193 scale=scale,