---
title: Dynamic Bayesian Updating
url: https://www.emergentmind.com/topics/dynamic-bayesian-updating
type: topic
---

# Dynamic Bayesian Updating

Dynamic Bayesian updating is the sequential revision of probabilistic beliefs as new evidence arrives, with the posterior from one update becoming the prior for the next. In the supplied literature, this idea appears in standard recursive Bayes, in generalized Bayesian or Gibbs posteriors when a full likelihood is unavailable or inconvenient, and in non-stationary settings where memory, forgetting, or adaptive transformation of prior information become part of the update mechanism. The updated object may be an activity-duration distribution, a latent graph trajectory, a dynamic linear model parameter, a distribution over markings in a condition/event net, a structural deterioration law, or a full probability measure over future observations [2605.17608] [2509.19748] [2202.02405] [2308.13069].

## 1. Recursive, generalized, and composite formulations

The canonical form of dynamic Bayesian updating is recursive conditioning. In the construction scheduling framework, the duration belief for activity \(i\) is updated from prior to posterior by
\[
P(D_i \mid O_i)=\frac{P(O_i \mid D_i)\,P(D_i)}{P(O_i)},
\]
and the posterior becomes the next prior as new observations arrive [2605.17608]. In dynamic linear models for portfolio selection, the same recursive structure appears in Kalman-filter-style updates,
\[
m_t=m_{t-1}+A_t(r_t-f_t), \qquad C_t=C_{t-1}-A_t Q_t A_t',
\]
so that \((\theta\mid D_t)\sim N(m_t,C_t)\) is the next period’s prior [1911.07526]. In the diachronic forecasting formulation, standard conditioning is written as
\[
P_n(x)=P_{n-1}(x\mid y_{n-1})=\frac{P_{n-1}(y_{n-1}x)}{P_{n-1}(y_{n-1})},
\]
with \(P_n\) a probability measure over the remaining future sequence [2308.13069].

A second formulation replaces a fully specified likelihood with a generalized Bayesian or Gibbs posterior. For dynamic random dot product graphs, the posterior is defined through a least-squares loss,
\[
\Pi_\lambda(X_{1:m}\mid Y_{1:m}) \propto \exp\!\left( -\lambda \sum_{t=1}^m \|Y_t - X_tX_t^\top\|_F^2 \right)\Pi(X_{1:m}),
\]
where \(\lambda>0\) is the learning rate. This posterior is the unique minimizer of a loss consisting of expected reconstruction error plus a Kullback–Leibler term relative to the prior, following Bissiri et al.’s generalized Bayes framework [2509.19748]. The same paper emphasizes that the learning rate controls how much the data dominate the prior, with \(\lambda\to 0\) returning essentially the prior and \(\lambda\to\infty\) concentrating around the empirical adjacency spectral embedding solution.

A third formulation concerns objective or noninformative priors when successive experiments relate to the same continuous parameter in different ways. Lewis argues that ordinary sequential updating can become order-dependent because the noninformative prior appropriate for experiment A may differ from the one appropriate for experiment B. The proposed correction is to treat the accumulated data as a composite experiment, multiplying likelihoods while adding Fisher information:
\[
p(\theta\mid x_A,x_B)\propto [L_A(\theta)L_B(\theta)]\,|h_A(\theta)+h_B(\theta)|^{1/2}.
\]
This replaces “posterior-to-prior updating” by “composite-experiment updating” and is explicitly motivated by order-invariance and probability matching [1308.2791]. This suggests that dynamic Bayesian updating is not always exhausted by repeated multiplication by a new likelihood; in some objective-Bayesian settings, the prior geometry must also evolve.

## 2. Evolving latent structure and state representation

The literature treats dynamic Bayesian updating as a mechanism for learning many different kinds of evolving latent structure. In construction digital twins, the project is represented as a directed acyclic graph
\[
G=(V,E),
\]
each activity has stochastic duration \(D_i\), and project completion time is the network-level random variable
\[
T=\max_{p\in P}\sum_{i\in p} D_i.
\]
Activity durations are modeled as lognormal random variables,
\[
D_i\sim \text{Lognormal}(\mu_i,\sigma_i),
\]
and are repeatedly revised as project evidence arrives [2605.17608].

In dynamic random dot product graphs, the latent state is geometric. A sequence of adjacency matrices \(\{Y_t\}_{t=1}^m\) is modeled through node positions \(x_{it}\in\mathbb{R}^d\) with
\[
\mathbb{E}(Y_{ij,t})=x_{it}^\top x_{jt}.
\]
Temporal dependence is introduced by Gaussian random walk priors. For \(r=1\),
\[
x_{it}=x_{i,t-1}+\sigma_i w_{it},
\]
and for \(r=2\),
\[
x_{it}=2x_{i,t-1}-x_{i,t-2}+\sigma_i w_{it}.
\]
Inference is only identifiable up to orthogonal transformations, so posterior contraction is stated for \(X_tX_t^\top\) and for \(X_t\) up to rotation [2509.19748].

Distributed monitoring problems use yet another state representation. In asynchronous dynamic Bayesian networks, each variable is a supernode containing subnodes \(X(t_0),X(t_1),\dots\), and continuous-time evolution is modeled with conditional intensity matrices \(Q[X\mid u]\). The transition over an interval \(\Delta t\) is
\[
P(X(t+\Delta t)\mid X(t),u)=\exp(Q[X\mid u]\Delta t).
\]
Beliefs are updated locally and asynchronously by message passing between supernodes and subnodes, rather than by globally synchronized time slices [1207.1398].

Complex dynamic Bayesian networks with non-Gaussian observations are represented cliquewise on junction trees. The state vector is partitioned as
\[
\theta_t=(\theta_t[1],\ldots,\theta_t[n_t]),
\]
and each observed block depends on a clique-local linear predictor
\[
\eta_t[j]=F_t[j]^\top \theta_t[j].
\]
This permits local approximate updates followed by exact Gaussian propagation on the junction tree [1301.6738].

Dynamic updating can also operate on symbolic hidden states. In condition/event nets, the hidden state is the marking \(m\subseteq S\), while the observer maintains a probability distribution over markings
\[
p:\mathcal{M}\to [0,1], \qquad \sum_{m\in\mathcal{M}}p(m)=1.
\]
Transition requests and observed success or failure induce updates to that distribution, and the corresponding modular Bayesian networks are manipulated by structural rewrites [1807.02566]. Taken together, these examples show that the “state” in dynamic Bayesian updating may be scalar, vector, path-valued, graph-valued, or combinatorial.

## 3. Evidence models and sequential assimilation

The evidence stream in dynamic Bayesian updating is frequently noisy, partial, and heterogeneous. In construction digital twins, the evidence includes BIM progress reports, drone imagery, IoT telemetry, computer vision tracking, productivity logs, site inspection reports, material delivery records, and weather/external condition data. The observation model is explicitly noisy:
\[
O_i = D_{\text{true}}+\varepsilon_i, \qquad \varepsilon_i\sim N(0,\sigma_{\text{obs}}).
\]
These observations are treated as imperfect signals about the underlying duration distribution rather than direct measurements of the true duration [2605.17608].

In condition/event nets, the evidence is not a numerical measurement but the outcome of an intervention: a requested transition either succeeds, fails due to precondition, or fails due to postcondition. The paper defines update operators \(ass_{A,b}\), \(nas_{A,b}\), and \(set_{A,b}\), and shows that
\[
success_t(p)=set_{\bullet t,1}\!\left(set_{t,0}\!\left(ass_{\bullet t,0}\!\left(ass_{t,1}(p)\right)\right)\right),
\]
while
\[
fail_t^{pre}(p)=nas_{t,1}(p), \qquad fail_t^{post}(p)=nas_{\bullet t,0}(p).
\]
These are proved to be Bayesian conditioning in disguise [1807.02566].

In embedding learning, the evidence arrives batch by batch. The triplet-sampling framework assumes that class-specific embeddings follow a multivariate Gaussian,
\[
x\sim \mathcal{N}(\mu^j,\Sigma^j),
\]
with a Normal-Inverse-Wishart prior on \((\mu,\Sigma)\). New mini-batches produce updated class means and covariances, and these updated distributions are then used to sample anchor-positive-negative triplets. The operational update uses posterior expectations for \(\mu^{0,j}\) and \(\Sigma^{0,j}\), making the process explicitly incremental over batches [2007.05610].

Clinical and structural monitoring problems assimilate evidence at coarser temporal resolutions. In the dynamic survival-prediction study, new data arrive by update period, and Bayesian updating uses the previous model’s coefficients as priors with a forgetting factor \(\xi<1\):
\[
\beta_{j,u}\sim N(\hat{\beta}_{j,u-1}, \hat{\sigma}_{j,u-1}/\xi).
\]
The process repeats across periods in rapidly changing environments [2305.00260]. In vibration-based structural health monitoring, dynamic response time series are processed by output-only operational modal analysis, specifically stochastic subspace identification, to extract modal quantities such as identified eigenvalues and mode shapes. These become the observed data in a sequential Bayesian model updating problem for deterioration parameters, and the updated posterior is propagated into structural reliability and hazard calculations [2103.07382]. This suggests that dynamic Bayesian updating is often a two-stage procedure: raw measurements are first converted into an informative intermediate representation, and only then assimilated probabilistically.

## 4. Exact updates, approximations, and computational machinery

Some dynamic Bayesian updates are analytically tractable because they use conjugate structure. The batch-incremental triplet method uses a Normal-Inverse-Wishart prior for a Gaussian likelihood, so the posterior remains Normal-Inverse-Wishart and the update is closed form. The paper further shows that the marginal posterior for the mean is multivariate Student-\(t\), while the marginal posterior for the covariance is inverse Wishart [2007.05610]. In dynamic generalized linear junction trees, local non-Gaussian observations are handled by choosing a conjugate family for the transformed latent variable \(\lambda_t[j]=g_j^{-1}(\eta_t[j])\), updating within that family, and then mapping back to a Gaussian approximation for \(\eta_t[j]\) by moment matching. The approximation quality is assessed with the Hellinger metric and theoretical bounds [1301.6738].

Other settings require approximate posterior computation because exact closed form is unavailable or computationally unattractive. In construction scheduling, evolving lognormal duration models with nonconjugate observation structure are updated by recursive maximum a posteriori estimation,
\[
\theta_{\text{post}}=\arg\max_{\theta_i}\left[\log P(O_i\mid \theta_i)+\log P(\theta_i)\right],
\]
with \(\theta_i=(\mu_i,\sigma_i)\) [2605.17608]. In multimodal structural model updating, the likelihood itself is approximated in latent space by a multimodal variational autoencoder. The trained surrogate unimodal encoders approximate \(p(z\mid X_{\text{obs}})\) and \(p(z\mid \theta)\), and posterior sampling is then carried out with Metropolis–Hastings using the VAE-based likelihood approximation [2406.09051].

Sampling-based computation is central in structural dynamics. Bayesian finite element model updating uses posterior sampling because the posterior is high-dimensional, nonlinear, and lacks a closed form. The compared methods are Metropolis-Hastings, Slice Sampling, and Hybrid Monte Carlo. The reported conclusion is that both MH and HMC perform better than Slice Sampling when Young’s modulus is chosen as the updating parameter, and that HMC gives better results than MH and Slice sampling techniques when the area moment of inertias and section areas are updated [1110.3382]. The more recent AM-SGHMC method augments SGHMC with neural networks that learn state-dependent diagonal \(G(z)\) and \(C(z)\) matrices, normalized energy and gradient inputs, and bounded-prior transformations. Because the inputs and outputs are designed to be scale-invariant across same-type structural problems, the trained sampler can be directly applied to various Bayesian updating problems of the same type of structure without further training [2604.25710].

Dynamic graph inference emphasizes scalability. The generalized Bayesian dynamic RDPG paper develops a Gibbs sampler whose precision matrix is banded and sparse, and reports that the cost of sampling latent positions scales as
\[
O(E),
\]
where \(E\) is the number of observed edges in the dynamic network, rather than \(O(mn^2)\) [2509.19748]. This suggests a general computational theme: dynamic Bayesian updating usually couples a recursive probabilistic rule with an approximation strategy tailored to the geometry, sparsity, or dimensionality of the latent state.

## 5. Non-stationarity, misspecification, and the limits of naive recursion

Several papers argue that naive recursive Bayes can fail in changing environments because it keeps accumulating evidence from outdated regimes. BAM states the problem directly: in non-stationary environments, a Bayesian never forgets. Its update rule introduces a memory buffer \(D\) and a readout weight vector
\[
W_t=[w_{t,1},\ldots,w_{t,t-1}],
\]
and defines
\[
p(\theta_t \mid \mathcal{D}_t, D, W_t) \propto p(\theta_t)\, p(\mathcal{D}_t \mid \theta_t)\, \prod_{j=1}^{t-1} p(\mathcal{D}_j \mid \theta_t)^{w_{t,j}}.
\]
With binary weights \(w_{t,j}\in\{0,1\}\), BAM chooses which past observations to remember and which to forget; it thereby generalizes recursive Bayes, power priors, exponential forgetting, and Bayesian unlearning [2202.02405].

Dynamic transformation of prior knowledge in streams takes a different route. TPS does not merely initialize a prior once; it repeatedly transforms external prior knowledge \(\eta\) into model parameters through
\[
\beta^t=f(\pi^t\eta), \qquad \pi_k^t \sim \mathcal{N}(\pi_k^{t-1}, \sigma I).
\]
The posterior at minibatch \(t\) is based on the transition prior \(p(\pi^t\mid \pi^{t-1})\) and the current minibatch likelihood under \(\beta^t\). The paper also states that some existing streaming approaches can forget initial knowledge at a rate of \(O(t^{-1})\), which motivates repeatedly re-encoding prior knowledge over time [2003.06123]. In the clinical survival-updating study, the forgetting factor \(\xi<1\) plays a related role by discounting older information when new data arrive [2305.00260].

Misspecification introduces a different limitation. For dependent data and possibly wrong models, posterior concentration is shown to occur around hypotheses with the smallest Kullback–Leibler divergence rate from the truth, not necessarily around a true model. The key dynamical assumption is a relative asymptotic equipartition property,
\[
\lim_{t\to\infty}\frac{1}{t}\log R_t(\theta)=-h(\theta) \qquad P\text{-a.s.},
\]
with sieve-like control on the prior ensuring that the posterior suppresses suboptimal hypotheses exponentially [0901.1342]. This suggests that in complex dynamic settings the practical target of Bayesian updating may be the best available approximation rather than exact recovery.

Adjacent literature also clarifies what dynamic Bayesian updating does not imply. In dynamic persuasion, delaying information provision does not affect the sender’s ex-ante value under divisible updating, and with a Bayesian receiver one-shot and two-step persuasion are equivalent in value. Strict gains or losses from gradual disclosure arise only when the receiver updates non-Bayesianly, as in Grether’s \(\alpha\)–\(\beta\) rule [2508.12328]. This is a useful contrast: temporal staging alone does not create a dynamic effect if the updating rule itself is dynamically consistent.

## 6. Forecasts, decisions, and evaluation of evolving beliefs

Dynamic Bayesian updating is typically used not for posterior estimation alone but for forecast generation, risk quantification, and decision support. In construction scheduling, posterior duration distributions are propagated by Monte Carlo simulation to produce
\[
E[T], \qquad \operatorname{Var}(T), \qquad P(T>T_{\text{target}}), \qquad CP_i=\frac{N_i^{\text{critical}}}{N}.
\]
These quantify expected completion time, completion-time variance, delay probability, and activity criticality [2605.17608]. In dynamic random dot product graphs, the generalized posterior predictive distribution supports \(k\)-step-ahead network forecasts and credible intervals for future edge probabilities [2509.19748]. In finance, updated AR or VAR parameter posteriors feed directly into dynamic mean-variance portfolio rules and optimal efficient portfolios under uncertain investment lifetime [1911.07526]. In clinical survival prediction, updated models are assessed out-of-sample with the C-index with inverse probability of censoring weighting, calibration intercept, calibration slope, and Brier score with IPCW [2305.00260]. In structural health monitoring, posterior deterioration parameters update structural reliability, hazard, repair timing, expected total life-cycle cost, value of information, and value of partial perfect information [2103.07382].

| Domain | Updated quantity | Reported outputs |
|---|---|---|
| Construction digital twins | Activity durations \(D_i\) | \(E[T]\), \(\operatorname{Var}(T)\), \(P(T>T_{\text{target}})\), \(CP_i\) |
| Dynamic random dot product graphs | \(X_{1:m}\), \(\sigma_{1:n}\), \(\lambda\) | \(k\)-step-ahead generalized posterior predictive distribution, credible intervals for future edge probabilities |
| Multi-period portfolio selection | AR or VAR parameters | Optimal efficient portfolios |
| Clinical survival prediction | Survival model updates across periods | C-index, calibration intercept, calibration slope, Brier score |
| Structural health monitoring | Deterioration parameters and reliability | Hazard, repair timing, VoI, VPPI |
| Time-updating forecast evaluation | Model bankroll or credibility | Market consensus probabilities, implied model credibilities |

The evaluation of evolving Bayesian beliefs is itself a subject of study. One line of work models forecast models as canonical Kelly bettors and interprets bankroll as Bayesian credibility. In the multinomial case, the update
\[
w_i'=\frac{p_i}{m_i}\sum_{k=1}^{n} m_k w_k
\]
is presented as a direct mathematical and conceptual analogue to Bayes’ theorem, with the ratio \(p_i/m_i\) playing the role of a likelihood ratio or Bayes factor contribution [2602.09982]. Another line develops game-theoretic betting protocols for a forecaster whose time-\(n\) belief is a probability measure over the remaining future, showing that game-theoretic test martingales are equivalent to visible measure-theoretic test martingales in finite settings and extending the framework to \(K\)-steps-ahead forecasting and nearly optimal sequential decision making [2308.13069]. This suggests that dynamic Bayesian updating is simultaneously an inference rule, a forecasting device, and an object of sequential calibration and betting-based scrutiny.

Across the supplied literature, the recurrent structure is a closed loop of prior, new evidence, posterior, forecast, and next update. In construction scheduling, this loop is written explicitly as “Prior \(\rightarrow\) Likelihood from new data \(\rightarrow\) Posterior \(\rightarrow\) Monte Carlo schedule simulation \(\rightarrow\) Updated risk metrics \(\rightarrow\) Next update” [2605.17608]. A plausible implication is that this loop is the most stable encyclopedia-level characterization of dynamic Bayesian updating across domains: it is Bayesian learning made explicitly temporal, repeatedly operationalized, and tied to current decisions rather than a single terminal inference.

Source: https://www.emergentmind.com/topics/dynamic-bayesian-updating