A Dirac-to-Weyl gamma-matrix similarity transform with the required signs is
It is a unitary matrix, with . Multiplication gives
The matrix commutes with every : their common block structure uses the same antisymmetric block matrix, and the scalar blocks of commute with each Pauli matrix. Hence , exactly the spatial matrices specified in the Weyl representation of the gamma matrices. A similarity transformation preserves every Clifford algebra relation, so no new consistency check is needed for the transformed representation.
Write every displayed block identity as . Direct block multiplication in the Dirac representation of the gamma matrices gives
Thus their mixed anticommutator is zero. For two spatial indices,
Using the Pauli matrix multiplication law gives
These exhaust all time-time, time-space and space-space cases of the Clifford algebra. No special choice of individual Pauli matrices beyond their stated anticommutator is needed.
The gamma matrices represent the spacetime Clifford algebra. With the Minkowski metric convention above,
The braces denote the anticommutator. Consequently , , and distinct gamma matrices anticommute. The identity matrix is essential: the right side is a matrix equality, not a scalar equality. These signs agree with the explicit Dirac representation of the gamma matrices in the next part.
The oscillator-generated covariant photon Fock space has an indefinite Hermitian form, not a positive Hilbert space inner product. For a normalizable one-photon wave packet of polarization , its squared norm is proportional to . Thus gives a negative-norm photon state. The divergent of an unsmeared momentum eigenstate is a separate normalization issue, avoided by the wave packet.
Choose contravariant polarization vectors , , and two with for . The third spatial polarization is longitudinal polarization; the first two are transverse polarization. The timelike photon polarization is distinct from the longitudinal one.
The Gupta-Bleuler quantization condition sets the divergence of the positive-frequency part of a quantum field to zero on physical states:
Restore the factors to the annihilation terms. Since and , Fourier transform gives the equivalent condition
Its sign depends on the chosen sign of the longitudinal polarization vector; the covariant condition does not.
To see its content for a general Fock state, temporarily discretize momentum and decompose one unphysical oscillator sector as , where . The temporal oscillator obeys , while . The condition therefore becomes
Equivalently, the allowed finite-particle states use the transverse creation operators and only in the unphysical sector. Indeed the constraint acts on a polynomial of the two unphysical creation operators as , whose kernel consists of polynomials in their difference. Also , and a state containing is orthogonal to every constrained state because annihilates every such state. This argument applies mode by mode and extends by smearing to continuum momentum.
For one photon, is constrained only when , producing a null state. The Gupta-Bleuler null-state quotient removes these null directions. The condition excludes negative-norm physical states; quotienting its null states leaves the two positive-norm transverse photon polarizations. The condition alone gives a positive semidefinite Hermitian form, not yet a positive definite Hilbert space.
Use the supplied oscillator expansion, with the boundary-term convention for momentum stated in part (b). Use a real polarization vector basis, as in the supplied unconjugated polarization completeness relation. Raise its second field index to obtain
The mixed commutator contains only the annihilation-creation and creation-annihilation terms. Their signs are both positive after combining the minus sign in the momentum expansion with the negative Minkowski metric oscillator commutator. Setting and using the momentum Dirac delta distribution gives
Here the last equality is the Fourier representation of the Dirac delta function. Similarly,
because the integrand is odd under . The momentum-momentum commutator is proportional to the same odd difference, now weighted by , and also vanishes. Therefore the equal-time canonical commutation relations are
These are identities of operator-valued distributions, understood after smearing. The extra indices in the TeX polarization relation are transcription defects; the PDF has the ordinary two-polarization completeness contraction used above. The direct momenta of part (b) also satisfy the canonical commutation relations after their boundary-generated shift, although they do not have the unmodified mode expansion printed here.
Write . The additional gauge fixing variation is . Integration by parts gives the Euler-Lagrange field equation , hence . This equation follows from the gauge-fixed density without imposing as a separate operator identity.
Direct differentiation of the density as printed gives the canonical momentum components conjugate to the lower-index fields:
There is a boundary-term convention to reconcile with part (c). The Feynman-gauge Maxwell kinetic density after a boundary-term subtraction is
It has the same Euler-Lagrange field equations, but its canonical momentum components are . The mode expansion supplied in part (c) is the expansion of these latter momenta. Thus it is valid after the stated boundary-term subtraction; it is not the direct derivative of the original density. For example, a time-independent spatially varying with gives but .
The boundary-term shift of canonical field momenta is a canonical transformation. If spatial boundary terms vanish, , and . This explains why the two conventions have the same dynamics while their momentum formulas differ.
Use natural units and the Minkowski metric throughout. An Abelian gauge transformation is . For a sufficiently regular , commuting the partial derivatives gives , so the Maxwell Lagrangian is invariant pointwise.
Vary the action with respect to the electromagnetic four-potential, using variations of compact support or vanishing on the boundary. The antisymmetry of the electromagnetic field tensor gives
After integration by parts, . The principle of stationary action therefore gives the source-free Maxwell equations,
In the Lorenz gauge, , this becomes with . The PDF's name “Lorentz gauge” denotes the usual Lorenz gauge, named after Lorenz. Residual gauge transformations preserve it when ; this condition is not itself a complete removal of the gauge freedom.
Let and . With conditional independence of the binary observations given the latent function, define
The prior distribution is specified on the precision , not on . Its Jacobian determinant gives
and the marginal posterior density is proportional to . Working instead with precision or logarithmic coordinates is possible, provided all corresponding Jacobian determinants are included.
Choose an importance sampling density covering the support of , draw independently from it, and use the nonnegative unbiased likelihood estimator
Each summand integrates to , proving unbiasedness. A simple always-valid choice is the latent Gaussian prior distribution, , when the covariance matrix is nonsingular. Then the estimator is just the average of the bounded values . For duplicate inputs, sample only the distinct latent values and multiply all likelihood contributions at each common input. A Gaussian approximation to the latent posterior distribution, or its mixture with the Gaussian prior distribution, can reduce relative variance while keeping complete support. Its approximation does not replace the exact weights.
The pseudo-marginal Metropolis–Hastings algorithm stores both the current and its estimator randomness ; write their sampling law as . Propose , independently generate , compute , and accept the pair with
Here denotes the proposal probability density function from to . On rejection retain both the old parameter and the old likelihood estimate. Initialize with a positive estimate. One must not independently refresh the denominator after rejection.
Correctness follows from the extended posterior density
The pair proposal is . Its Metropolis–Hastings acceptance probability simplifies to the displayed ratio, since the auxiliary densities cancel. Hence detailed balance holds for . Integrating over gives by unbiasedness, so the parameter marginal distribution is exactly the intended posterior distribution for any fixed positive . This is exactness of the stationary target, not independent exact samples or a guarantee of rapid convergence. Large relative variance of the likelihood estimates can produce long holding times, so marginalizing does not by itself guarantee an improvement in mixing time.
The most plausible cause is strong dependence between the latent variables and covariance hyperparameters in their joint posterior distribution. In the usual conditionally independent Gaussian process classification model, put , , and . The probit regression likelihood is , while has a multivariate normal distribution with covariance matrix .
For distinct inputs and positive length scales, is positive definite. Its Gaussian probability density function contributes
The unit-rate exponential prior distribution on therefore gives
Its scale is tightly tied to the current magnitude and shape of . Changing changes the eigenvectors and eigenvalues of its covariance matrix; a latent vector typical under the old covariance may be very atypical under a substantially different one. Holding fixed can thus restrict the range of a hyperparameter update, and holding the hyperparameters fixed restricts the next latent-function update. Even exact sampling of each full conditional distribution can move slowly along a narrow ridge of the joint posterior distribution.
Binary observations provide limited information about the absolute size of large correctly signed latent values. This can leave substantial scale uncertainty and strengthen the dependence. Long correlation lengths can also make the covariance matrix nearly singular and the latent coordinates highly dependent. These are plausible explanations for a large mixing time, not proofs that every data set causes slow convergence. A blocked Gibbs sampler removes dependence within its chosen blocks, but not dependence between them. Repeated identical inputs should be represented by one shared latent value; otherwise the nominal Gaussian matrix is singular and the displayed inverse formula must be replaced by a representation on its support.
Parameterize the initial product normal distribution by , using
Run the given recursion with independent random variables with uniform distributions to obtain . Because the noise laws do not depend on , the chain rule and automatic differentiation give reparameterization gradients through the entire simulated trajectory. The variational objective is , where and .
One must account for the output differential entropy, not just optimize . Simulation and differentiability of do not themselves supply an evaluable . If retains the reversible Markov chain assumption from part (b), a concrete algorithm avoids a transition-density oracle using reversible density-ratio propagation. Write , with an evaluable unnormalized posterior density, and define
Detailed balance implies
For example, integrate and replace by to obtain the first identity. This also proves absolute continuity of measures of with respect to when . An everywhere finite differentiable log-posterior has positive density, so an ordinary product normal distribution has this domination.
For each outer sample , independently run trajectories of length starting at , with endpoints . These are reverse trajectories because of detailed balance; the same recursion simulates them. Estimate by
Average over an outer batch. Compute it stably by a logarithmic sum of exponentials. Differentiate through both outer and reverse trajectories and through the explicit product-normal probability density function . Only the given state gradients of and , together with elementary normal-density derivatives, are needed. Apply stochastic gradient descent to and , refreshing the independent noise batches. The logarithmic scale parameters keep all variances positive.
This is nested Monte Carlo optimization of the exact objective in the limit of adequately increasing inner and outer sample sizes. For fixed , is generally biased, even though is an unbiased estimator. Finite logarithmic moments, suitable uniform integrability, and justified differentiation under the integral sign are needed for consistency and gradient interchange. Under those conditions increasing and controlling the optimization error yields a justified numerical approximation; nonconvexity prevents a generic guarantee of a global optimum.
There is a real assumption issue: part (c) redefines without explicitly restating reversibility. The algorithm above uses the reversible interpretation inherited from part (b). For an arbitrary differentiable recursion, one additionally needs access to output densities or a simulatable reverse kernel and regularity of the objective. Differentiability alone is insufficient: is differentiable and, with a standard-normal target, makes a point mass for , so every candidate has infinite Kullback-Leibler divergence. Thus a universal finite-objective algorithm cannot be justified solely from the literal differentiability assumptions. An invertible recursion with computable density transformations offers a different sufficient route, but invertibility is not stated here.
A counterexample can use an irreducible, aperiodic reversible Markov chain, so failure is not caused by a pathological transition structure. Take two binary coordinates, enumerate the four states as , and let be their uniform distribution. Use the Markov kernel
Its rows sum to one and it is symmetric, proving detailed balance with . Every entry is positive. Let the univariate parametric family be .
For a product initial probability distribution with parameters , additivity of the Kullback-Leibler divergence gives
Each summand has derivative on , so the unique optimum at has . Thus
On the other hand the allowed initial probability distribution , obtained with , is sent by one step to . Its Kullback-Leibler divergence is zero, the smallest possible value. Consequently
The transition reduces Kullback-Leibler divergence for each input, but may reduce it by different amounts for different inputs. It therefore need not preserve the ordering of competing initial probability distributions, which is exactly why Markov chain variational inference must optimize the initial parameters for its chosen number of steps.
Write . Mean-field variational inference minimizes the reverse Kullback-Leibler divergence
Each is a probability density function; without further parametric restrictions, the optimization is over all such product probability distributions. Equivalently it maximizes the evidence lower bound for any unnormalized posterior density . Its difference from is exactly the Kullback-Leibler divergence.
Fix and write
Assume and that the displayed expected values and objective decomposition are well defined. The optimal factor is
Indeed, the part of the objective depending on is
The remaining term is fixed. Gibbs inequality gives nonnegativity of the Kullback-Leibler divergence, with equality precisely at almost everywhere. This proves global optimality of that coordinate update; it does not assert global optimality of a sequence of coordinate updates for the full nonconvex product-family problem.
The printed upper index in the list of remaining coordinates is inconsistent with the dimension ; the natural interpretation is . Also, if the exponential expression has zero or infinite normalizing constant, or the expected values are undefined, the usual coordinate formula needs additional support or integrability hypotheses. A restricted parametric factor family need not contain this unrestricted optimal factor.
Take observation periods to have positive lengths , and let be the Markov kernel of a sweep which first draws given the current and then draws the new vector . Write . Conditional on , independence and the gamma distribution mean and variance give
Averaging over the full conditional distribution of preserves this bound, uniformly in the current state. Therefore . This is the mechanism in bounded conditional moments imply a geometric drift.
Choose , , and . Outside , ; inside , . Hence
For completeness, is a small set, even though it approaches the boundary of the positive orthant. Choose small enough that . On , . The conditional gamma distribution of has fixed shape and rate in a compact positive interval. Its probability density function therefore has a positive common lower bound on . For those , the product probability density function of similarly has a positive common lower bound on , since every shape is positive and every rate lies in a compact positive interval. Integrating over gives on , where is the uniform distribution on and .
The transition probability density function is positive throughout the positive orthant, giving an irreducible Markov chain. Since , the box minorization includes starts in the same box, which gives an aperiodic Markov chain. The proper prior distributions and bounded Poisson distribution likelihood have positive finite evidence, so the invariant posterior distribution is proper. Together with the geometric drift condition, these hypotheses imply geometric ergodicity. In particular, the argument proves the required drift rather than assuming that every Gibbs sampler is geometrically ergodic.
The positive-period convention matters. If zero lengths are allowed, the printed assertion can fail: take , , . Then and , so
This example has a proper posterior distribution but cannot satisfy a finite quadratic geometric drift condition. Thus the proof establishes the intended claim for positive observation periods, not an unrestricted extension to zero exposure.
Use the shape-rate convention for every gamma distribution, with . Let and . The Poisson distribution likelihood and the Gamma–Poisson hierarchical model give
All factors involving the coordinate being updated must be retained, including from the conditional gamma distributions. The full conditional distributions are
For a blocked Gibbs sampler, draw from the second full conditional distribution, then redraw all the independently from the first ones. Repeating these two steps preserves the joint posterior distribution; observing after each complete sweep gives the Markov chain used in the next part. Standard gamma distribution sampling works for noninteger as well as integer shapes. The positive observation-period convention is used for the subsequent geometric drift condition.
An exact finite mixture model follows from the binomial theorem:
Define positive mixture weights
Here is the Beta function, so the component probability distributions are . All coefficients are computable from the observed integer counts.
Use inverse transform sampling with the supplied to choose : writing and , take . Then recycle its position inside that interval,
Recycling a uniform random variable after discrete sampling makes uniform independently of , since . It is also independent of all supplied random variables with gamma distributions. Transform these by the probability integral transform:
They are independent random variables with uniform distributions because has a unit-rate exponential distribution.
Conditional on , put and . Form the values and return their th order statistic. This uses at most the supplied random variables with gamma distributions, since . The uniform order statistic formula gives
Averaging these component probability density functions with weights recovers exactly the normalized posterior density. Unused inputs can be discarded. When , , , and the output is , as required. Interval endpoints and ties have probability zero; any consistent convention there is harmless. Exactness here concerns the mathematical algorithm; finite-precision arithmetic can be stabilized with logarithmic mixture weights.
Adaptive rejection sampling is exact rejection sampling for a differentiable log-concave probability density. Given interior evaluation points , construct the upper envelope
where is the unnormalized target probability density function. The concave function lies below all its tangent lines, so . Interpolation between adjacent evaluation points gives a lower envelope there, allowing a squeeze test. The normalized upper envelope is a piecewise exponential proposal distribution.
Draw and an independent from the uniform distribution, and accept when . A squeeze test can establish acceptance without evaluating . Otherwise evaluate and, whenever it is evaluated, add to the envelope's points. Repeat until acceptance. On each linear piece , its integral is on , with value when . These integrals select a piece; inverse transform sampling samples within it.
Here, with and , the log-posterior on has
Choose evaluation points strictly inside , avoiding any endpoint singularities of the logarithm. Every finite set of tangent lines gives a finite upper-envelope integral on this bounded interval; slope conditions needed for unbounded supports are unnecessary. When the posterior distribution is simply the uniform distribution and may be sampled directly.
For any fixed envelope, proposal times acceptance is proportional to . Conditional on all past evaluations and rejections, the next accepted value therefore has the same normalized posterior density. To see that adaptation does not spoil this, at every attempted draw its accepted mass is a scalar times that fixed posterior density; summing over possible rejection histories preserves it. The acceptance probability stays bounded below by its positive initial-envelope value because insertion only lowers the upper envelope. Thus acceptance occurs almost surely. With fresh independent randomness, successive accepted values are exact independent samples from the posterior distribution, even when the envelope is retained and improved between them.
Put and . Dropping factors independent of , the multinomial likelihood and uniform prior give the posterior density
Multiplying this by the stipulated binomial distribution for the latent variable cancels the factor :
Consequently the two full conditional distributions are
Start with , sample from its binomial distribution, then sample a new from its Beta distribution, and repeat. The binomial distribution can be sampled by adding independent Bernoulli random variables. For the Beta distribution, take independent and and return ; integer-shape gamma distributions are sums of independent unit-rate exponential distributions. Both shapes are positive even when some observed counts vanish.
Each Gibbs sampler update preserves the augmented posterior distribution, so its marginal distribution is the required posterior distribution. The positive full conditional distributions on the interior allow exploration of the entire support. These iterates are generally dependent; they are not the independent exact draws constructed in the later parts.
The reference kernel printed in the hint has total mass , so it is not a Markov kernel. We explicitly replace it by the normalized random transposition shuffle: choose two positions independently and uniformly and swap them. Its probabilities are
These sum to one, and its ordinary spectral gap is , the value intended in the hint. This value can also be certified by the Aldous spectral gap theorem: the single-label generator has rate between distinct positions, so on centered functions it acts as multiplication by . Both shuffles have the same uniform distribution on a finite set as stationary distribution.
We use the Canonical paths comparison theorem in the following precise form. For two finite reversible Markov chains with the same positive stationary distribution , route each directed transition of along a positive-capacity path of . If
then and . Indeed, telescope each difference along its path, apply the Cauchy-Schwarz inequality, and sum with weights . The load definition gives the Dirichlet form of a Markov chain inequality; the common variance denominator then gives the spectral gap inequality. Identity transitions use empty paths.
Put . A transposition is routed using the word
of length . Each generator occurs at most twice; write its occurrence count as . For a fixed directed adjacent edge and each occurrence of in a word, there is exactly one starting permutation whose translated word crosses this edge at that occurrence. This follows because right multiplication by the prefix is a bijection of the symmetric group. Hence the uniform stationary factors cancel in the comparison load, leaving
The crude bounds , , and imply . Therefore
Only pairs with contribute, and there are of these. Keeping this restriction gives the stronger and . Neither comparison bound needs the exact adjacent-shuffle spectral gap; the normalization repair of the reference kernel is essential.
The symmetric group has elements, not the printed order . For the identity and the adjacent transpositions have probabilities , so the specified transition matrix is normalized. Its stationary distribution is uniform and it satisfies detailed balance, since each transposition is its own inverse. Adjacent transpositions generate the symmetric group, making this an irreducible Markov chain; the positive holding probability makes it an aperiodic Markov chain.
For a fixed label , right multiplication swaps positions, so is the label's position. Under the uniform distribution on a finite set it is uniform on . Therefore
For each generator , the squared change is one if the label occupies either of the two positions, and zero otherwise. Its stationary expected value is , giving
Thus the tagged-label Rayleigh quotient for adjacent transpositions gives the corrected bound
The printed factor six cannot be obtained: the hint's variance of a uniform distribution is wrong. Directly, . Bounded convergence in distribution of does imply convergence of its first two moments, hence convergence of its variance, but the limit is .
More strongly, the claimed asymptotic bound itself is false for this kernel. The Aldous spectral gap theorem states that the interchange process on a finite connected weighted graph has the same spectral gap as the single-label random walk with identical edge rates; see Theorem 1.1 and its proof. Here is that interchange process generator on a path graph with edge rates . For the tagged-label generator, direct substitution of gives eigenvalues , , including the reflecting endpoint equations. These distinct eigenvalues exhaust the tagged-label function space. Consequently the theorem yields
which exceeds asymptotically. This use of a substantial external theorem is only to certify the printed claim's failure; the corrected factor-twelve upper bound above follows directly from the requested test function and the variational formula.

Pinned article: Introduction to the OurBigBook Project

Welcome to the OurBigBook Project! Our goal is to create the perfect publishing platform for STEM subjects, and get university-level students to write the best free STEM tutorials ever.
Everyone is welcome to create an account and play with the site: ourbigbook.com/go/register. We belive that students themselves can write amazing tutorials, but teachers are welcome too. You can write about anything you want, it doesn't have to be STEM or even educational. Silly test content is very welcome and you won't be penalized in any way. Just keep it legal!
We have two killer features:
  1. topics: topics group articles by different users with the same title, e.g. here is the topic for the "Fundamental Theorem of Calculus" ourbigbook.com/go/topic/fundamental-theorem-of-calculus
    Articles of different users are sorted by upvote within each article page. This feature is a bit like:
    • a Wikipedia where each user can have their own version of each article
    • a Q&A website like Stack Overflow, where multiple people can give their views on a given topic, and the best ones are sorted by upvote. Except you don't need to wait for someone to ask first, and any topic goes, no matter how narrow or broad
    This feature makes it possible for readers to find better explanations of any topic created by other writers. And it allows writers to create an explanation in a place that readers might actually find it.
    Figure 1.
    Screenshot of the "Derivative" topic page
    . View it live at: ourbigbook.com/go/topic/derivative
  2. local editing: you can store all your personal knowledge base content locally in a plaintext markup format that can be edited locally and published either:
    This way you can be sure that even if OurBigBook.com were to go down one day (which we have no plans to do as it is quite cheap to host!), your content will still be perfectly readable as a static site.
    Figure 5. . You can also edit articles on the Web editor without installing anything locally.
    Video 3.
    Edit locally and publish demo
    . Source. This shows editing OurBigBook Markup and publishing it using the Visual Studio Code extension.
  3. https://raw.githubusercontent.com/ourbigbook/ourbigbook-media/master/feature/x/hilbert-space-arrow.png
  4. Infinitely deep tables of contents:
    Figure 6.
    Dynamic article tree with infinitely deep table of contents
    .
    Descendant pages can also show up as toplevel e.g.: ourbigbook.com/cirosantilli/chordate-subclade
All our software is open source and hosted at: github.com/ourbigbook/ourbigbook
Further documentation can be found at: docs.ourbigbook.com
Feel free to reach our to us for any help or suggestions: docs.ourbigbook.com/#contact