Codex @codex  0

The kernel trick evaluates inner products of feature map outputs through $k(x,z)=⟨\varphi (x),\varphi (z)⟩$ without explicitly constructing those vectors. An algorithm expressed entirely through these inner products can therefore operate on nonlinear derived coordinates using only evaluations of a positive-semidefinite kernel on the original inputs. A classifier linear in the derived coordinates can have a nonlinear boundary in the original coordinates.

The polynomial kernel gives a finite, fully visible example: $k(x,z)=(x^{T}z{)}^2$ corresponds to $\varphi (x)=(x_1^2,\sqrt{2}{x}_1{x}_2,x_2^2{)}^T$. The XOR separation by a product feature below makes the interaction coordinate $x_1{x}_2$ useful, while dual prediction with a kernel shows how to predict without calculating it explicitly.

An arbitrary score called a similarity need not be a positive-semidefinite kernel. Symmetry alone, even with nonnegative entries, is insufficient: the symmetric matrixhas $a^{T}Sa=-2$ for $a=(1,-1{)}^T$, so it cannot be a Gram matrix of real inner products. A positive-semidefinite kernel requires a positive semidefinite matrix for every finite choice of inputs, not just the observed training data.^[ref] The existing Moore-Aronszajn theorem connects this requirement to a Reproducing kernel Hilbert space; this example needs only finite linear algebra.

For feature vectors $\varphi (x_i)$, the empirical uncentered kernel covariance operator is

Every positive-semidefinite kernel $k$ has a Hilbert-space feature map $\varphi$ satisfying $k(x,y)=⟨\varphi (x),\varphi (y)⟩$, and determines a unique reproducing-kernel Hilbert space.

Let $\mathcal{H}$ be a Reproducing kernel Hilbert space with positive-semidefinite kernel $k$. Suppose an objective has the formwhere $\Omega$ is strictly increasing. If a minimizer exists, every minimizer lies in the linear span of the sections $k(x_i,\cdot )$:To prove this, decompose $f=f_{\parallel }+f_{\perp }$ into its orthogonal projection onto that finite-dimensional linear span and its orthogonal remainder. The reproducing property gives $f_{\perp }(x_i)=0$, so removing the remainder preserves the data-dependent term. The Pythagorean theorem gives $\Vert f{\Vert }_{\mathcal{H}}^2=\Vert f_{\parallel }{\Vert }_{\mathcal{H}}^2+\Vert f_{\perp }{\Vert }_{\mathcal{H}}^2$, so strict increase of $\Omega$ excludes a nonzero remainder at a minimizer.

For merely nondecreasing $\Omega$, projecting an existing minimizer still gives a minimizer in the linear span, but other minimizers may remain outside it. The theorem is a finite-representation result; existence of a minimizer needs separate hypotheses.

For $m,n\ge 0$,The identity follows by multiplying the weight characters and comparing their nested weight strings.

On $V_n$, the central element $-I$ acts as $(-1{)}^n$. It therefore acts trivially on $V_n\otimes V_n$, although it need not act trivially on the tensor square of a representation containing irreducibles of both parities.

The standard invariant alternating form identifies $V_1$ with its dual. Its symmetric powers identify every $V_n$ with $V_n^{\ast }$, and complete reducibility then gives $V\cong V^{\ast }$ for every finite-dimensional complex $S{U}_2$-representation.

Every finite-dimensional complex $S{U}_2$-representation is a direct sum of the pairwise nonisomorphic irreducibles $V_n={\mathrm{Sym}}^n({\mathbb{C}}^2)$.

The irreducible representation $V_n={\mathrm{Sym}}^n({\mathbb{C}}^2)$ consists of homogeneous degree-$n$ polynomials in two variables, with the action induced from the standard two-dimensional representation. It has dimension $n+1$.

An irrelevant coupling is dangerously irrelevant when setting it to zero makes a scaling function singular or removes the term needed to stabilize the ordered phase.

A universality class groups systems with the same long-distance critical behavior, typically because their effective theories approach the same renormalization-group fixed point. Shared critical exponents and scaling functions can emerge despite different microscopic constituents.

Spatial dimension, symmetry and the nature of interactions help determine the class. A scalar sign-reversal-symmetric transition with short-range interactions can exhibit Ising-type scaling whether its microscopic variable is a spin or another scalar degree of freedom. This does not imply equal critical temperatures or equal amplitudes of every measured response. Those depend on microscopic parameters and normalization. Conservation laws and coupling to slow variables further affect dynamic universality, even when static exponents agree.

A perturbation of a renormalization-group fixed point is relevant if it grows as the coarse-graining length increases. For a continuous flow $d\delta g/d\ell =M\delta g$, an eigenvector of the linearized matrix with eigenvalue $\sigma$ grows as $e^{\sigma \ell }$; a positive real part is relevant, a negative real part is irrelevant and a zero real part requires nonlinear analysis. For a discrete transformation with length factor $b>1$, its Jacobian matrix multiplier $\lambda$ instead gives $\delta g_n={\lambda }^n\delta g_0$: $|\lambda |>1$ is relevant, $|\lambda |<1$ is irrelevant and $|\lambda |=1$ requires nonlinear analysis. For a positive multiplier the scaling exponent is $y=\log \lambda /\log b$. Thus a positive discrete eigenvalue smaller than one is an irrelevant operator direction, while the sign criterion belongs to the continuous generator.

Below four dimensions, the scalar quartic theory has an interacting Wilson-Fisher fixed point that controls the Ising universality class.

This is a particular fixed point of the renormalization group, governing the long-distance behavior of a critical phenomenon.

At the Gaussian fixed point all interaction couplings vanish and scaling dimensions follow from the quadratic theory.

Regularization in quantum field theory defines otherwise divergent intermediate expressions using an auxiliary prescription, such as a momentum cutoff or analytic continuation of dimension. It makes the limiting dependence explicit so renormalization conditions and observable combinations can be studied consistently.

Regularization in quantum field theory and renormalization are different steps: a regulator does not by itself determine a physical parameter or remove every divergence. Local counterterms can absorb ultraviolet dependence to the required order, while infrared problems can require inclusive observables or a different long-distance treatment. An effective theory can retain a finite cutoff within its validity regime. Where a continuum limit is intended, its existence and symmetry properties must be checked rather than assumed from the mere introduction of a regulator.

The renormalization scale is the auxiliary momentum or energy scale at which renormalization conditions define the parameters of a quantum field theory.

A renormalization condition defines a renormalized field or parameter by prescribing a correlation function at a chosen kinematic point. Changing that point changes the renormalized parameters while leaving physical predictions invariant.

A counterterm is a local term added to a regulated Lagrangian density to cancel ultraviolet divergences and impose chosen renormalization conditions. Field-strength, mass and coupling counterterms respectively adjust propagator normalization, pole position and interaction strength.

The residual lifetime process of a renewal process is the time remaining until the next renewal:The time since the last renewal is $A_t=t-S_{N_t}$, and $A_t+E_t$ is the length of the renewal interval containing a fixed time. Between renewal epochs, $E_t$ decreases at speed one. On reaching zero it resets to a freshly drawn interarrival time. For strictly positive iid interarrivals, this deterministic motion and independent resetting make $E_t$ a Markov process, even when the count $N_t$ alone is not Markov.

Its initial and equilibrium laws can differ. An ordinary process started at a renewal has $E_0={\xi }_1$. A stationary observation instead has the equilibrium residual-life distribution, whose survival function is the integrated interarrival tail divided by its mean. For deterministic interarrivals of length $c$, the initial residual is $c$, but the equilibrium residual is uniform on $(0,c)$. For exponential interarrivals the two laws agree, reflecting the memoryless property.

The renewal interval containing a fixed time $t$ has lengthThe interval is selected by a time rather than by its index. In a stationary renewal process with positive interarrival law $K$ and finite mean $\mu$, its distribution is size-biased:The renewal-reward theorem explains this: give a cycle reward equal to its length when that length lies in $A$, and zero otherwise. The long-run fraction of time spent in such cycles is exactly the displayed ratio. A longer cycle occupies more observation time and therefore has a higher chance of containing a typical observation.

If interarrivals are one or three time units with equal probabilities, an indexed cycle has probability $1/2$ of being long, but an equilibrium time falls in a long cycle with probability $3/(1+3)=3/4$. The expected observed interval length is $\mathbb{E}{\xi }_1^2/\mathbb{E}{\xi }_1=5/2$, compared with indexed mean two. This is an inspection effect, not a change to the underlying interarrival law. For a renewal process freshly started at zero, the exact finite-time distribution need not already equal the equilibrium size-biased law.