---
title: 'SMC–WFR: Weighted Forward Recursion'
url: https://www.emergentmind.com/topics/sequential-monte-carlo-smc-wfr
type: topic
---

# SMC–WFR: Weighted Forward Recursion

Sequential Monte Carlo (SMC–WFR) encompasses a family of methodologies within sequential Monte Carlo that utilize weighted forward recursions, advanced resampling algorithms, or weight-function resets to achieve variance control, robustness to path degeneracy, and efficiency in high-dimensional or rare-event inference. The abbreviation “SMC–WFR” is variously used in the literature for “Weighted Forward Recursion” [1012.5390], “Weight-Function Resetting” [1903.04797], and “Weighted Finite Resampling” [1903.12583]. Recent developments also connect SMC–WFR to Wasserstein–Fisher–Rao (WFR) gradient flows for optimization in measure spaces [2506.05905]. SMC–WFR enables online smoothing, sharp asymptotic control of estimator variance, and, in specialized forms, logarithmically efficient importance sampling for large-deviation problems [1202.4582]. This article provides a comprehensive account of SMC–WFR from formal models and algorithms to advanced theoretical properties and variants.

## 1. State-Space Setting and Smoothing via Weighted Forward Recursion

Consider a hidden Markov model (HMM) on a state-space $\mathcal{X}$ with observations $Y_k$ in $\mathcal{Y}$. The joint density, under parameter $\theta$, has the form
\[
p_\theta(x_{0:n},y_{0:n}) = \mu_\theta(x_0)\prod_{k=1}^n f_\theta(x_k|x_{k-1}) \prod_{k=0}^n g_\theta(y_k|x_k)
\]
with $X_0\sim \mu_\theta(\cdot)$, $X_k|X_{k-1}=x_{k-1}\sim f_\theta(\cdot|x_{k-1})$, and $Y_k|X_k=x_k\sim g_\theta(\cdot|x_k)$.

The central inferential objective is the recursive computation of smoothed expectations of additive path functionals:
\[
S_n(x_{0:n}) = \sum_{k=1}^n s_k(x_{k-1},x_k), \qquad \mathcal{S}_n^\theta = \mathbb{E}_\theta\left[ S_n(X_{0:n}) \mid y_{0:n} \right]
\]
A forward-smoothing (weighted forward recursion, WFR) enables one to compute $\mathcal{S}_n^\theta$ recursively in $n$ without path storage, through auxiliary functions
\[
T_n^\theta(x_n) = \mathbb{E}_\theta\left[ S_n(X_{0:n}) \mid y_{0:n-1}, X_n=x_n \right]
\]
satisfying
\[
T_n^\theta(x_n) = \int \left[ T_{n-1}^\theta(x_{n-1}) + s_n(x_{n-1}, x_n) \right] p_\theta(x_{n-1}|y_{0:n-1}, x_n) dx_{n-1}
\]
with $T_0^\theta(x_0)\equiv 0$ [1012.5390].

## 2. The SMC–WFR Algorithm: Structure, Recursion, and Pseudocode

In practice, the filtering distribution $p_\theta(x_n|y_{0:n})$ and backward kernel $p_\theta(x_{n-1}|y_{0:n-1}, x_n)$ are not available in closed form. SMC–WFR substitutes empirical measures constructed from a weighted particle cloud:
- At time $n-1$, maintain particles $\{(X_{n-1}^j, W_{n-1}^j)\}_{j=1}^N$ and corresponding estimates $\{\widehat{T}_{n-1}(X_{n-1}^j)\}$.
- Propagate and reweight to obtain $\{(X_n^i, W_n^i)\}_{i=1}^N$ approximating $p_\theta(dx_n|y_{0:n})$.
- For each $i$, update the forward-smoothing estimate by
\[
\widehat{T}_n(X_n^i) = \frac{\sum_{j=1}^N W_{n-1}^j f_\theta(X_n^i|X_{n-1}^j)\left[ \widehat{T}_{n-1}(X_{n-1}^j) + s_n(X_{n-1}^j, X_n^i) \right]}{\sum_{j=1}^N W_{n-1}^j f_\theta(X_n^i|X_{n-1}^j)}
\]
- The smoothed additive expectation is then
\[
\widehat{\mathcal{S}}_n = \sum_{i=1}^N W_n^i \widehat{T}_n(X_n^i)
\]
This results in a fully online algorithm, summarized as follows:

```
Algorithm SMC–WFR
INITIALIZE n=0: for i=1..N, sample X_0^i∼μ_θ(·), set w₀^i=g_θ(y₀|X₀^i), normalize W₀^i, set ẐT₀^i←0
FOR n=1,2,… do
  1) Resample ancestors, Propagate: Xₙ^i∼f_θ(·|X_{n−1}^{aₙ(i)})
  2) Weight: wₙ^i ← g_θ(yₙ|Xₙ^i), normalize Wₙ^i
  3) For i=1..N, update ẐTₙ^i via weighted sum over all j=1..N (as above)
  4) Ŝₙ=∑_{i=1}^N Wₙ^i * ẐTₙ^i
end
```
Under this scheme, the key smoothing weights are
\[
\widetilde{w}_{n-1}^{j|i} \propto W_{n-1}^j f_\theta(X_n^i|X_{n-1}^j), \quad \sum_{j=1}^N \widetilde{w}_{n-1}^{j|i}=1
\]
[1012.5390].

## 3. Theoretical Properties: Variance, Consistency, and Central Limit Theorems

SMC–WFR is designed to control the variance of smoothed estimators and mitigate path degeneracy:
- **Mean-square error**: For $\|s_k\|, \operatorname{osc}(s_k) \leq 1$ and all $n$,
\[
\mathbb{E}\left|\widehat{\mathcal{S}}_n - \mathcal{S}_n\right|^2 \leq C\frac{n+1}{N}\left(1+\sqrt{\frac{n+1}{N}}\right)^2
\]
where $C$ is independent of $n,N$.
- **Central Limit Theorem**: As $N\to\infty$,
\[
\sqrt{N}\left(\widehat{\mathcal{S}}_n - \mathcal{S}_n\right) \xrightarrow{d} \mathcal{N}(0, \sigma_n^2)
\]
with $\sigma_n^2$ growing at most linearly with $n$. This is in marked contrast with path-space particle smoothers, for which the variance increases at least quadratically in $n$ [1012.5390].
- The analysis is based on telescoping the estimation error into one-step increments, using contractive properties of the forward and backward kernels and Khinchine inequalities for each step.

SMC–WFR is proven to be **unbiased** for functions of the terminal state and establishes **consistency** and **CLT** properties for a wide function class, with variance bounded by the (at most) linear-in-time accumulation due to resampling noise [2208.12108].

## 4. Algorithmic Variants: Weight-Function Resetting and Matrix-Resampling

The WFR principle applies beyond forward-smoothing. In SMC with Weight-Function Resetting [1903.04797], blocks of SMC resample and normalize weights, controlling early path degeneracy by periodically setting incremental weights to unity:
- Fix reset times $1 = \rho_0 < \rho_1 < \cdots < \rho_K = T$, either uniformly or adaptively (e.g., ESS-triggered).
- On block $[\rho_{k-1}+1,\rho_k]$ run standard SMC, initializing and resetting weights at the start of each block.

Table: Distinct SMC–WFR Algorithms and Their Key Elements

| Variant                | Main Mechanism           | Noted Effect/Strength                     |
|------------------------|-------------------------|-------------------------------------------|
| Weighted Forward Rec.  | Forward additive update | Linear-in-time variance for smoothing     |
| Weight-Function Reset  | Blockwise weight resets | Controls path degeneracy, robust Z-est.   |
| Weighted Finite/Residual| Residual resampling    | Variance reduction over multinomial       |

Matrix-resampling [1903.12583] encapsulates WFR as “residual” resampling and identifies optimal schemes:
- Weighted finite (residual): allocate integer copies by floor of normalized weights, then use multinomial allocation for fractional parts.
- Stratified (sorted): sorts by an informative statistic, then stratifies, achieving minimal resampling variance for that statistic.

## 5. Extensions: Logarithmic Efficiency, Random-Weight SMC, and Rare-Event Estimation

WFR schemes are central to rare-event simulation and SMC with unbiased random weights:
- **Logarithmic Efficiency**: For estimation of small probabilities (e.g., rare events in Markov additive processes), SMC–WFR resampling weights are constructed to mimic the incremental likelihood ratio of an optimal exponential tilt. This achieves a variance-to-square mean ratio of order $\exp(o(n))$ for event probabilities $P_n \sim \exp(-nI)$ [1202.4582].
- **Random-Weight SMC**: The estimator is consistent in probability under minimal $L^1$ moment conditions, and a CLT holds for functionals in a recursively defined $L^2$ function class; resampling ensures the variance is a sum over steps, rather than a multiplicative product, preventing exponential variance growth [2208.12108].

Such SMC–WFR estimators play a critical role in high-dimensional and rare-event simulations, where classical importance sampling would otherwise fail due to poor proposal adaptation or intractable optimal tilting.

## 6. SMC–WFR in Wasserstein–Fisher–Rao Gradient Flows

Recent work [2506.05905] establishes a distinct but nomenclature-overlapping SMC–WFR method to approximate Wasserstein–Fisher–Rao gradient flows in measure space optimization:
- The WFR distance on $\mathcal{P}(\mathbb{R}^d)$ interpolates between 2-Wasserstein (mass-conserving) and Fisher–Rao (mass-varying) transports.
- The WFR gradient flow of $D_{KL}(\cdot\|\pi)$ for sampling is discretized by splitting into a Fokker–Planck (Wasserstein) step and a pure Fisher–Rao (replicator) step.
- Each iteration computes an unadjusted Langevin move (diffusion) followed by importance re-weighting, then resampling, corresponding to the forward-only SMC paradigm.
- The method achieves exponential-rate KL contraction under log-Sobolev assumptions, and numerically outperforms competing schemes on multimodal targets.

## 7. Computational and Practical Considerations

- **Complexity**: The basic SMC–WFR (forward-smoothing) update involves $O(N^2)$ computation per timestep, reducible to $O(N\log N)$ with specialized summation/data structures [1012.5390].
- **Memory**: Only current and previous particle clouds and associated forward variables are required, yielding $O(N)$ storage overhead.
- **Parallelism and Implementation**: SMC–WFR variants such as blockwise reset and chain-autocovariance-based variance estimation are naturally parallelizable and suitable for modern architectures [2011.02328, 1903.04797].
- **Variance Control**: All WFR-family methods are constructed to suppress variance, either by summary-weight propagation (forward recursion), tailored resampling (matrix schemes), blockwise resets, or use of rare-event tilting information.

## References

- "Forward Smoothing using Sequential Monte Carlo" [1012.5390]
- "Elements of Sequential Monte Carlo" [1903.04797]
- "Unifying Sequential Monte Carlo with Resampling Matrices" [1903.12583]
- "Convergence of random-weight sequential Monte Carlo methods" [2208.12108]
- "A sequential Monte Carlo approach to computing tail probabilities in stochastic models" [1202.4582]
- "Sequential Monte Carlo approximations of Wasserstein--Fisher--Rao gradient flows" [2506.05905]
- "Waste-free Sequential Monte Carlo" [2011.02328]

Source: https://www.emergentmind.com/topics/sequential-monte-carlo-smc-wfr