Scott Sheffield Created 2026-10-13 Updated 2026-11-05
Scott Sheffield authored probability teaching material on conditional expectation, random walks and martingales. Conditional expectation supplies the common language for predictions given partial information and for processes whose current state is the best conditional prediction of a later value.[ref]
18.175 Lecture 16: Conditional expectation, random walks, martingales by Scott Sheffield. author attribution and the described mathematical development
Portrait of Scott Sheffield
. Via Wikimedia Commons; credits and usage information are on the linked source file page. Source. Selection on an extreme first measurement 2026-10-14
Select units with first reading , keep that membership fixed, and measure the same units again. The first reading now has a truncated normal distribution. For a general finite threshold , write for the standard normal density and for the standard normal distribution function. Since ,The tower property of conditional expectation lets us average the line over the selected group.[ref] At this givesThe average second reading is below the selection threshold even though every selected first reading exceeded it. This is a group expectation; any finite generated group's sample mean fluctuates around it. Selecting again on would define a different group and answer a different question.
The latent variable decomposition makes the apparent drop especially clear. Gaussian conditioning gives and . ThusThe first measurement error has zero unconditional mean, but positive selected mean. The second measurement error stays centered because it is independent of selection on . The persistent signal never changed.
Simple kriging Created 2026-10-05 Updated 2026-10-18
Simple kriging assumes a known mean function. If and are known, center the observations, predict the residual, and add back the target mean:For known zero mean, these reduce to and . Completing the covariance quadratic form giveswhich proves minimum error variance among predictors of the form . No sum-to-one constraint is needed: the known mean already supplies the offset. In the exponential example from ordinary kriging, imposing known zero mean gives weights , prediction , and error variance . The smaller error variance uses additional knowledge of the mean; it is not a free improvement when the mean is unknown.
If has a joint multivariate normal distribution, the displayed prediction is exactly , and is its conditional variance. A conditional expectation minimizes mean squared error among all square-integrable functions of the observations. Without a joint multivariate normal distribution, simple kriging remains the best predictor in the specified affine class, but the conditional expectation may be nonlinear and its conditional variance may depend on . Thus Gaussian process regression and the second-moment kriging construction agree under these assumptions, while having different starting points.[ref]
State estimation Created 2026-10-03 Updated 2026-10-13
State estimation infers an unobserved state of a dynamical system from its observations and a model of its evolution. An observed position can help estimate an unobserved velocity; repeated noisy observations can help distinguish changes in the state from measurement noise. The result should describe both an estimate and its uncertainty. A Kalman filter does this exactly for a linear Gaussian state-space model, maintaining a conditional expectation and conditional covariance.
The time at which information is available matters. Causal state filtering estimates the present state from observations available now. State prediction estimates a future state from those same observations. Later observations can improve an estimate of a past state, but that is a different information set from the one a live controller had. Kalman explicitly distinguishes these estimation tasks in A New Approach to Linear Filtering and Prediction Problems (1960), section Optimal Estimates.
Stochastic gradient second-moment identity 2026-10-25
For an unbiased stochastic gradient at with finite conditional second moment, expand in the squared Euclidean norm. The cross term has zero conditional expectation, which proves the displayed identity. It separates the squared true gradient from the total conditional variance of its random estimate.
Target: gradient-descent.bigb, under Stochastic gradient descent. Stuart P. Lloyd 2026-11-05
Stuart P. Lloyd developed conditions and an iterative procedure for minimizing squared error in quantization. His paper Least Squares Quantization in PCM by Stuart P. Lloyd, published in 1982, identifies Bell Laboratories as his affiliation and records a presentation of part of the work in 1957. The motivating problem was pulse-code modulation: representing a signal sample by one of finitely many reconstruction values.[ref]
Least Squares Quantization in PCM by Stuart P. Lloyd. Page 129, author footnote and introduction
The central conditions have a clear interpretation. For fixed reconstruction values, assign an input to a nearest value. For fixed assignment regions of positive probability, place each reconstruction value at the corresponding conditional expectation. Alternating these operations gives Lloyd's algorithm; neither operation increases the mean squared error, but this does not guarantee the global optimum from every initialization.[ref]
Least Squares Quantization in PCM by Stuart P. Lloyd. Sections IV–VI: best quanta, best partition and an algorithm
For an original example, take a continuous uniform distribution on and two reconstruction values. Splitting at zero gives conditional means and , whose midpoint is again zero. The mean squared error isUsing the single reconstruction value zero instead gives . This concrete improvement explains the optimization objective: preserve signal amplitudes accurately with a limited collection of values.
Sub-sigma-algebra Created 2026-10-25 Updated 2026-10-31
A sigma-algebra on the same underlying set as , with . In probability theory it represents information retained from the original probability space and is a possible conditioning collection for a conditional expectation.
Supermartingale Created 2026-09-24 Updated 2026-11-05
A supermartingale is an integrable adapted process satisfying almost surely. Its conditional expectation can decrease, although a realized sample path need not be decreasing. For a mean-zero random walk , the process with is a supermartingale: the conditional expected next change is . Negating a supermartingale gives a submartingale. A nonnegative supermartingale converges almost surely to a finite limit, but its limiting mean may be smaller than its earlier means without uniform integrability.
Symmetrization as conditional expectation 2026-10-14
If a finite group acts by measure-preserving maps on a probability space, let be its invariant sigma-algebra. For an integrable random variable , the displayed group average is -measurable. For , changing variables under each group element gives . The average therefore satisfies the defining integral identity for conditional expectation. For permutations of exchangeable coordinates, averaging a product attached to distinct indices yields the sum over all ordered distinct -tuples divided by the falling factorial .
Tensorization of a Poincaré inequality Created 2026-09-28 Updated 2026-11-05
If each probability measure has Poincaré constant , then the product measure has Poincaré constant at most . This follows by iterating the law of total variance and applying Jensen inequality to derivatives of conditional expectations.
Unbiased stochastic gradient Created 2026-09-29 Updated 2026-11-05
A random vector is an unbiased stochastic gradient at an iterate ifFor an adaptive iteration it is often useful to require the stronger historical condition , where contains all information available before drawing the new sample. This states explicitly which information is held fixed in the conditional expectation.
For , a conditionally uniform index gives . If index instead has positive sampling probability , then is the correctly weighted estimate. Omitting these factors generally changes its expectation.
Uniform integrability of conditional expectations Created 2026-10-05 Updated 2026-11-05
For one integrable random variable , the family , over arbitrary sigma-algebras , has uniform integrability. On , the defining property of conditional expectation gives , while by the Markov inequality. The integrable random variable has uniformly small absolute integrals over events of sufficiently small Probability.
Variance tensorization Created 2026-10-03 Updated 2026-10-18
Write for all inputs except . Under the Efron–Stein inequality hypotheses, variance tensorization is the equivalent conditional variance formConditioning freezes the other inputs. Then are independent with the same conditional probability distribution, givingTaking expected values explains the factor .[ref]
Lecture 7a: Efron Stein inequality by Purnamrita Sarkar. Lecture 7a: Efron Stein inequality, slides 2 and 7, PDF pp. 3 and 8: conditional variances and the independent-copy identity
For Bernoulli input with parameter , let be the value of with coordinate set to one minus its value with that coordinate set to zero, keeping fixed. The two-point conditional variance isFor the maximum example, and . The two averaged conditional variances are and , precisely half the respective resampling terms. In the dependent example , each other coordinate already determines , so both conditional variances vanish; variance tensorization also fails without independence.
For a short induction proof, separate the last input using the law of total variance. Apply the smaller-dimensional statement to the conditional expectation; Jensen's inequality for the square bounds its coordinate contributions by the original ones. One input gives equality, starting the induction.[ref]
Lecture 7a: Efron Stein inequality by Purnamrita Sarkar. Lecture 7a: Efron Stein inequality, slides 3–6, PDF pp. 4–7: variance decomposition and conditional Jensen step; the induction outline here reorganizes those ingredients

