Multiplicative noise (carrier-scaled random input)¶
A random input changes an observation or state through an effect whose size depends on the signal or state carrying it.
Core Idea¶
Multiplicative noise here means that the size of a random input's effect depends on the signal or state it acts through. A coherent-image model multiplies reflectance by a random speckle factor.[1] A stochastic state model instead uses a noise coefficient that depends on the current state.[2] The first describes an observation and the second an evolving equation; they share an input rule rather than a distribution or a method.
Four questions identify the bounded relation. What is the underlying carrier? What is the random driver? How does the driver act through a carrier-dependent factor or coefficient? What observation or transition results? Remove carrier dependence, and the remaining disturbance can be additive.[1][2]
Both cases contain a Function Mapping as an internal constituent: the image product uses the carrier-to-amplitude rule \(g(x)=x\) before the separate random factor acts, while the Landau–Stuart example uses the carrier-to-coefficient rule \(g(x)=\sigma x\) before stochastic forcing acts.[1][2] The mapping alone is not the four-role noise relation; the random driver and receiving observation or transition remain necessary.
Scope of Application¶
Bioucas-Dias and Figueiredo give a source-described coherent-imaging case. Their reflectance image is observed through a speckle factor, with one-look exponential and multilook Gamma forms under their stated assumptions.[1] In their unit-mean product model \(Y=XN\), the departure from the unaltered signal is \(X(N-1)\).[1] The effect scales with \(X\), but this does not say that the factor \(N\) changes its own law with \(X\).[1]
Pavliotis gives an unlike formal case: a fluctuating parameter in a Landau–Stuart equation produces forcing with a coefficient proportional to the current state.[2] That written white-noise equation needs a stated Itô or Stratonovich interpretation for a unique stochastic process.[2] It is a worked model, not a measured deployment or the image paper's product observation.[2]
The examples support this carrier-dependent input rule only. They do not establish a universal claim about photographic defects, financial dynamics, particular diffusions or simulation practice.
Clarity¶
“Signal-dependent” describes the random effect, not necessarily the factor's own probability distribution. If a factor follows the same declared law at different reflectance values, the product disturbance can still change magnitude with reflectance.[1] In the Landau–Stuart example the coefficient multiplying formal noise changes with state.[2] Treating the image factor as an SDE increment would confuse two different model objects.
The relation is also distinct from a description of varying conditional variance. Heteroscedasticity may arise under additional assumptions, but a variance pattern alone does not identify the random input and carrier-dependent rule. A constant-scale additive disturbance is a negative case even though it remains random. Pavliotis names the constant state-independent coefficient additive.[2]
Manages Complexity¶
The four-role test separates the common mechanism from model-specific consequences. The imaging paper's speckle distribution, signal-to-noise calculation and logarithmic estimation method depend on its own assumptions.[1] The state equation's stochastic interpretation and drift belong to the second setting.[2] Neither set of details defines all multiplicative noise.
Abstract Reasoning¶
For a proposed case, identify the random input and write its connection to the carrier. If replacing that connection with a constant coefficient leaves the same asserted identity, the case has not demonstrated this bounded multiplicative relation. Next identify whether the result is a measured observation or a state transition, and declare the law or calculus convention needed for the particular quantitative claim.
The cited sources do not establish a universal log transform, lognormal output, fixed signal-to-noise ratio or one stochastic convention for all cases. The paper's log-domain method and Pavliotis's interpretation discussion are case-specific.[1][2]
Knowledge Transfer¶
The input-rule questions transfer between the image and state models, while their factor laws, estimators and stochastic calculus do not. The one-time image product does not require repeated positive compounding, so it is not automatically Multiplicative Random Growth. The nonlinear Landau–Stuart example is not the particular linear Itô model of Geometric Brownian Motion.[1][2] Any further setting needs its own model and source check.
Relationships to Other Abstractions¶
Current abstraction Multiplicative noise (carrier-scaled random input) Domain-specific
Parents (1) — outgoing typed relations
-
Multiplicative noise (carrier-scaled random input) includes Function (Mapping) Prime
The carrier-dependent random-input rule contains a deterministic Function Mapping from carrier value to the input amplitude or coefficient.
Hierarchy path (1) — routes to 1 parentless root
- Multiplicative noise (carrier-scaled random input) → includes Function (Mapping)
Neighborhood in Abstraction Space¶
Multiplicative noise (carrier-scaled random input) sits in a sparse region of the domain-specific corpus (71st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2769 abstractions)
Nearest neighbors
- Dynamic light scattering — 0.84
- Kushner–Stratonovich Equation — 0.84
- Computational Lagrangian Particle Tracking — 0.83
- Scattering — 0.83
- Score (statistics) — 0.83
Computed from structural-signature embeddings · 2026-10-25
References¶
[1] José M. Bioucas-Dias and Mário A. T. Figueiredo, “Multiplicative Noise Removal Using Variable Splitting and Constrained Optimization”, IEEE Transactions on Image Processing 19(7), 2010, DOI 10.1109/TIP.2010.2045029, printed pp. 1720–1722, §§I.A–B and II, especially equations (1)–(4). Primary research paper; its speckle law and estimation method are case-specific. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j
[2] Grigorios A. Pavliotis, Stochastic Processes and Applications, author manuscript dated 11 November 2015, §3.1 printed pp. 49–50, especially Example 3.1 and equations (3.1)–(3.4); §3.5 printed pp. 65–66; §3.6 printed pp. 68–69. Author-worked formal stochastic models, not an observed deployment. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k