# `IsGaussian.lean` -- Informal Summary > **Source**: [`GaussianField/IsGaussian.lean`](../../GaussianField/IsGaussian.lean) > **Generated**: 2026-03-03 11:30 > **Note**: Auto-generated by `/lean-summarize`. Re-run to refresh. ## Overview Shows that the Gaussian measure constructed by this library satisfies Mathlib's `ProbabilityTheory.IsGaussian` typeclass. The key bridging argument is that every continuous linear functional $L$ on $\text{WeakDual}\,\mathbb{R}\,E$ (the weak-* topology) is evaluation at some $f \in E$: continuity in the initial topology implies $\ker L$ contains an intersection of finitely many evaluation kernels, so by linear algebra $L$ is a finite linear combination of evaluations. Once this is established, the existing `pairing_is_gaussian` immediately yields the result. ## Status **Main result**: Fully proven None -- file is sorry-free. **Length**: 160 lines, 0 definition(s) + 2 theorem(s)/lemma(s) --- ### [`weakDual_clm_eq_eval`](../../GaussianField/IsGaussian.lean#L55) -- Lemma **Statement**: Every continuous linear functional $L : (\text{WeakDual}\,\mathbb{R}\,E) \to_L \mathbb{R}$ is evaluation at some $f \in E$: there exists $f$ such that $L(\omega) = \omega(f)$ for all $\omega$. **Proof uses**: `mem_span_of_iInf_ker_le_ker`, `Submodule.mem_span_range_iff_exists_fun`, `topDualPairing` --- ### [`measure_isGaussian`](../../GaussianField/IsGaussian.lean#L142) -- Instance **Lean signature** ```lean instance measure_isGaussian (T : E →L[ℝ] H) : ProbabilityTheory.IsGaussian (measure T) ``` **Informal**: The measure constructed from $T$ is Gaussian in the Mathlib sense: its pushforward by every continuous linear functional is a real Gaussian. **Proof uses**: [`weakDual_clm_eq_eval`](../../GaussianField/IsGaussian.lean#L55), [`pairing_is_gaussian`](../../GaussianField/Properties.lean), [`measure_centered`](../../GaussianField/Properties.lean), [`second_moment_eq_covariance`](../../GaussianField/Properties.lean), `variance_of_integral_eq_zero` --- *This file has **0** definitions and **2** theorems/lemmas (0 with sorry).*