論文の概要: Approximation Families and the Statistical Efficiency of Denoising Score Matching
- arxiv url: http://arxiv.org/abs/2610.36205v1
- Date: Mon, 26 Oct 2026 19:41:25 GMT
- ステータス: 翻訳完了
- システム内更新日: 2026-10-28 21:04:31.425897
- Title: Approximation Families and the Statistical Efficiency of Denoising Score Matching
- Title(参考訳): スコアマッチングの近似と統計的効率
- Abstract要約: 前方拡散の固有関数の勾配によって生成される有限次元近似空間上でのスコアマッチングについて検討した。
一次元オルンシュタイン-ウレンベック流下の近似とトーラス上の熱流下の三角-ポリノミカル近似との比較を行った。
- 参考スコア(独自算出の注目度): 0.0
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: We study score matching over finite-dimensional approximation spaces generated by gradients of eigenfunctions of the infinitesimal generator of a forward diffusion. We compare polynomial approximation under one-dimensional Ornstein-Uhlenbeck flow with trigonometric-polynomial approximation under heat flow on the torus. In the Ornstein-Uhlenbeck setting, we show that for every non-Gaussian initial distribution uniformly comparable to the standard Gaussian, the asymptotic score-estimation variance grows exponentially along a subsequence of polynomial degrees. Averaging the score-matching loss under the diffusion semigroup, which we call operator-informed score matching (OISM), reduces the variance at each fixed degree but does not eliminate this exponential instability at sufficiently small times. A local minimax lower bound shows that the same high-degree dependence arises when estimating the polynomial projection of the score, rather than being an artifact of OISM. In contrast, under heat flow on the torus, OISM over trigonometric polynomials admits finite-sample control of the score error along the forward diffusion and yields the classical minimax rate for estimating sufficiently smooth densities under expected Kullback-Leibler loss.
- Abstract(参考訳): 前方拡散の無限小発生器の固有関数勾配によって生成される有限次元近似空間上のスコアマッチングについて検討した。
一次元オルンシュタイン-ウレンベック流下の多項式近似とトーラス上の熱流下の三角-ポリノミカル近似との比較を行った。
Ornstein-Uhlenbeck 設定では、ガウスの初期分布が標準ガウス分布と一様に等しい場合、漸近的なスコア推定の分散は多項式次数列に沿って指数関数的に増加することを示す。
演算子インフォームドスコアマッチング (OISM) と呼ばれる拡散半群下でのスコアマッチング損失の平均化は, 一定度毎の分散を減少させるが, この指数的不安定性を十分に小さくは排除しない。
局所ミニマックス下限は、OISMの人工物ではなく、スコアの多項式射影を推定する際に同じ高次依存が生じることを示している。
対照的に、トーラス上の熱流下では、三角多項式上のOISMは、前方拡散に沿ったスコア誤差の有限サンプル制御を認め、期待されるクルバック・リーバーの損失の下で十分に滑らかな密度を推定するための古典的なミニマックス速度を得る。
関連論文リスト
- Poisson empirical Bayes estimation of sums of random variables via minimum-distance methods [6.4980926971596675]
可観測変数と可観測変数の関数の和の推定は、統計学における長年の問題である。
我々は、正規および粗大化最小距離推定に基づく非パラメトリック実証ベイズ手法を開発した。
そのような和の幅広いクラスに対して、その結果のプラグイン推定がベイズ推定とマージすることを示す大きなサンプル保証を確立する。
論文 参考訳(メタデータ) (2026-10-05T18:08:04Z) - Connecting Score Matching, Maximum Likelihood, and Expectation-Maximization in Mixed Linear Regression [11.223656960922478]
混合線形回帰(MLR)における応答の分散保存拡散と未知の混合重みについて検討した。
本分析では,不規則な拡散雑音レベルにおける損失幾何と最適化信号から,スコアマッチングの統計的保証を分離する。
論文 参考訳(メタデータ) (2026-09-04T19:52:26Z) - Minimax Optimality of Score-Entropy Discrete Diffusion [4.831510571088151]
具体的スコア推定の基本的な統計的限界について検討する。
スコアエントロピー損失の下での下位境界に一致するMLEに基づくしきい値推定器を提案する。
対象分布に対して、この密度比は、均一かつマスキングな離散拡散モデルの下で自然に制御されていることを示す。
論文 参考訳(メタデータ) (2026-08-21T00:16:00Z) - Wasserstein bounds for denoising diffusion probabilistic models via the Föllmer process [0.0]
本稿では,2-ワッサーシュタイン距離における拡散確率モデル(DDPM)のサンプリング誤差境界について検討する。
我々は, 段数と段数の両方で最適である鋭い上界を確立し, 文献で得られたいくつかの鋭い誤差境界を復元する。
本分析は,従来の逆Ornstein-Uhlenbeck法ではなく,Fllmer法を離散化したものである。
論文 参考訳(メタデータ) (2026-05-18T08:52:09Z) - Equilibrium Propagation Without Limits [0.0]
我々は,nudged と free phase のヘルムホルツ自由エネルギーの差の勾配が,まさに期待される局所エネルギー微分の差であることを示した。
これは、古典的なContrastive Hebbian Learning Updateを、任意の有限ヌーディングのための正確な勾配推定器として検証する。
論文 参考訳(メタデータ) (2025-11-27T01:55:26Z) - Structural aspects of FRG in quantum tunnelling computations [68.8204255655161]
一次元の4次元高調波発振器とダブルウェルポテンシャルの両方を探索する。
ポテンシャルV_k(varphi)と波動関数再正規化Z_k(varphi)の2つの偏微分方程式について検討した。
論文 参考訳(メタデータ) (2022-06-14T15:23:25Z) - Nonconvex Stochastic Scaled-Gradient Descent and Generalized Eigenvector
Problems [98.34292831923335]
オンライン相関解析の問題から,emphStochastic Scaled-Gradient Descent (SSD)アルゴリズムを提案する。
我々はこれらのアイデアをオンライン相関解析に適用し、局所収束率を正規性に比例した最適な1時間スケールのアルゴリズムを初めて導いた。
論文 参考訳(メタデータ) (2021-12-29T18:46:52Z) - Spectral clustering under degree heterogeneity: a case for the random
walk Laplacian [83.79286663107845]
本稿では,ランダムウォークラプラシアンを用いたグラフスペクトル埋め込みが,ノード次数に対して完全に補正されたベクトル表現を生成することを示す。
次数補正ブロックモデルの特別な場合、埋め込みはK個の異なる点に集中し、コミュニティを表す。
論文 参考訳(メタデータ) (2021-05-03T16:36:27Z) - On Linear Stochastic Approximation: Fine-grained Polyak-Ruppert and
Non-Asymptotic Concentration [115.1954841020189]
The inequality and non-asymptotic properties of approximation procedure with Polyak-Ruppert averaging。
一定のステップサイズと無限大となる反復数を持つ平均的反復数に対する中心極限定理(CLT)を証明する。
論文 参考訳(メタデータ) (2020-04-09T17:54:18Z) - Minimax Optimal Estimation of KL Divergence for Continuous Distributions [56.29748742084386]
Kullback-Leibler の同一および独立に分布するサンプルからの発散は、様々な領域において重要な問題である。
単純で効果的な推定器の1つは、これらのサンプル間の近辺 k に基づいている。
論文 参考訳(メタデータ) (2020-02-26T16:37:37Z)
関連論文リストは本サイト内にある論文のタイトル・アブストラクトから自動的に作成しています。
指定された論文の情報です。
本サイトの運営者は本サイト(すべての情報・翻訳含む)の品質を保証せず、本サイト(すべての情報・翻訳含む)を使用して発生したあらゆる結果について一切の責任を負いません。