A state space is part of every random-variable definition
Objective: Without looking back, derive “If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses.,” identify the controlling norm or domain, reproduce the checkpoint, and name the first failed line under “A collection of measurable coordinates indexed by an uncountable nonseparable system need not define one strongly measurable H-valued state.”
Unit 1, chapter 1 begins with the concrete question “A state space is part of every random-variable definition.” An infinite-dimensional state is one mathematical object in a declared space, not an instruction to write infinitely many coordinates. Before calculating, identify what can be observed, which finite projections are being used, and which conclusion concerns the full state rather than only those projections.
Read the definition in ordinary language first: An H-valued random variable is a measurable map X from a probability space into a declared separable Hilbert space H. Every displayed symbol must name its source and target. A vector, scalar random variable, covariance operator, cylindrical coordinate rule, Hilbert–Schmidt noise map, unbounded generator, and stochastic process are different object types even when the same letter is traditionally reused.
The bounded theorem target is: If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses. The conclusion belongs to the stated Hilbert spaces, filtration, time interval, operator domain, covariance convention, and moment class. Coordinatewise truth, weak truth after testing, norm convergence, almost-sure equality, and equality in law are not interchangeable.
The proof route is to start from “An H-valued random variable is a measurable map X from a probability space into a declared separable Hilbert space H.”; type the state space, noise space, operator domain, sigma-field, and integrability class; establish only the estimate named in “If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses.”; and stop at “A collection of measurable coordinates indexed by an uncountable nonseparable system need not define one strongly measurable H-valued state.”. Typical valid arrows use Parseval’s identity, monotone convergence, the Itô isometry, a Hilbert–Schmidt norm estimate, semigroup bounds, a fixed-point contraction, Galerkin compactness, monotonicity, coercivity, or Gronwall’s inequality. Naming a theorem without checking its operator and measurability hypotheses proves nothing.
The finite reconstruction is: In H=R^3, X takes (1,0,-1) and (-1,2,1) with probabilities 1/4 and 3/4; list its two possible states and verify total probability. Keep basis labels, operator directions, eigenvalues, time steps, units, and exact fractions visible. The audited endpoint is The state set is {(1,0,-1),(-1,2,1)} and 1/4+3/4=1. A finite projection can check algebra and an error budget; by itself it cannot certify the omitted tail or a full infinite-dimensional limit.
The failure mutation is part of the lesson: A collection of measurable coordinates indexed by an uncountable nonseparable system need not define one strongly measurable H-valued state. Find the first quantity that becomes undefined, infinite, nonmeasurable, nonpredictable, or outside the generator domain. Preserve all earlier valid lines and state the extra hypothesis required for repair instead of silently applying a finite-dimensional formula.
Keep six evidence layers separate. Geometry specifies norms and operator ideals. Probability specifies laws and filtrations. Stochastic integration controls accumulated noise. Semigroup or variational analysis gives solution meaning. Approximation controls discarded modes and time error. Modeling identifies what an equation represents. Success at one layer never certifies the other five.
Close the chapter by reconstruction, not recognition. Rebuild the typed definition, justify the decisive estimate, reproduce the finite endpoint by two routes, activate the mutation, and report the strongest surviving conclusion. Familiar notation, a smooth-looking simulation, or one stable coordinate plot is not mastery.
Keep a three-column audit beside every derivation. Column one names the full infinite-dimensional object and its legal space. Column two records the finite projection actually calculated. Column three records the theorem and quantitative tail estimate that reconnect the projection to the full object. If column three is empty, report only the finite result. If an operator domain, trace, stochastic-integral norm, or compactness step is missing, mark that exact gap before continuing.
Bounded theorem: If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses. No stronger topology, regularity, solution concept, or application claim is included.
Type every object in “An H-valued random variable is a measurable map X from a probability space into a declared separable Hilbert space H..” For each map write source, target, domain, measurability, adaptedness, and whether its norm is operator, trace, Hilbert–Schmidt, Bochner Lp, or path supremum norm.
Choose one orthonormal basis only as a computational instrument. Write every basis sum as a limit of finite partial sums, identify the nonnegative or square-summable budget that controls the tail, and prove that the final object does not depend on the chosen basis.
Check the probability layer. Verify strong measurability or the stated weak substitute, predictability before integration, covariance positivity, and the required first or second moment. A formal Gaussian coordinate family is not automatically an H-valued random variable.
Establish the analytic budget in “If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses..” Bound the relevant trace, Hilbert–Schmidt norm, semigroup convolution, coercive term, or Lipschitz constant on the declared interval before passing to a limit.
Apply the decisive arrow: start from “An H-valued random variable is a measurable map X from a probability space into a declared separable Hilbert space H.”; type the state space, noise space, operator domain, sigma-field, and integrability class; establish only the estimate named in “If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses.”; and stop at “A collection of measurable coordinates indexed by an uncountable nonseparable system need not define one strongly measurable H-valued state.”. If the result is mild, retain the semigroup and stochastic convolution; if variational, retain the V–H–V* pairing; if strong, prove domain membership rather than inferring it from notation.
Reconstruct the checkpoint independently: In H=R^3, X takes (1,0,-1) and (-1,2,1) with probabilities 1/4 and 3/4; list its two possible states and verify total probability. Verify The state set is {(1,0,-1),(-1,2,1)} and 1/4+3/4=1. Then repeat using a second route such as an eigenbasis sum, a trace identity, scalar testing, an energy equality, or a discrete recursion.
Activate “A collection of measurable coordinates indexed by an uncountable nonseparable system need not define one strongly measurable H-valued state.” and name the first failed line. The preceding steps prove exactly “If X is strongly measurable, each scalar coordinate <X,h> is measurable; for separable H, measurable coordinate approximations can reconstruct strong measurability under the stated hypotheses.”; they do not prove Banach-space stochastic integration, singular-noise renormalization, sharp regularity, production discretization accuracy, parameter identifiability, or application validity.
Finite model to reconstruct: In H=R^3, X takes (1,0,-1) and (-1,2,1) with probabilities 1/4 and 3/4; list its two possible states and verify total probability. Preserve basis labels, operator directions, exact values, and the omitted-tail record.
- Write the complete finite input for “In H=R^3, X takes (1,0,-1) and (-1,2,1) with probabilities 1/4 and 3/4; list its two possible states and verify total probability.”: spaces, basis vectors, retained modes, eigenvalues, operator matrices, initial state, time interval, and every probability or step-size convention.
- Compute each retained coordinate separately, then rebuild the vector or operator. Label whether the sum represents a norm, trace, covariance, stochastic integral, convolution, drift, or approximation error.
- Apply the chapter estimate while keeping exact arithmetic. Display the tail or residual term instead of replacing it by an unexplained zero.
- Recompute the endpoint by an independent scalar test, trace calculation, energy balance, or recurrence. Check object types and units as well as numerical equality.
- Both valid routes reach The state set is {(1,0,-1),(-1,2,1)} and 1/4+3/4=1. State the hypothesis that made them agree, activate the boundary mutation, and record the first unavailable norm or domain statement.
Result: Audited endpoint: The state set is {(1,0,-1),(-1,2,1)} and 1/4+3/4=1. This certifies only the stated finite checkpoint under the declared assumptions.