In a balanced category with pullbacks in a category, pull a monomorphism back across the bottom map of a lifting square whose left map is an epimorphism. The resulting monic projection is epic because its composite with the lifted top map is epic. Balancedness makes that projection invertible, producing the unique lift. Conversely, if all monomorphisms are strong monomorphisms, applying the lifting property of a bimorphism to its own square gives its inverse, so the category is balanced. Pullback stability of epimorphisms is not needed.

The logarithmic scoring rule rewards the natural logarithm of the reported probability mass or probability density at the observed outcome. The title uses the reward convention. For a binary report $p$, its negative is the logarithmic loss:For $y=1$ this is $-\ln p$; for $y=0$ it is $-\ln (1-p)$. At an endpoint, use these outcome-specific forms: assigning zero probability to the outcome incurs infinite loss, while assigning it probability one gives zero. The natural logarithm therefore makes a confidently wrong binary forecast far more costly than the bounded Brier score does.

Under the stipulated Bernoulli distribution with $q=0.7$, the expected logarithmic loss isIts unique minimum is at $p=0.7$, where the expected loss is approximately $0.610864$. At $p=0.2$ it is approximately $1.193550$. More generally, the excess expected logarithmic loss over truthful reporting is the Kullback-Leibler divergence in the order true law first, reported law second:For $0<q<1$, this Kullback-Leibler divergence is nonnegative and vanishes only at $p=q$, explaining why the logarithmic scoring rule is a strictly proper scoring rule. The same Kullback-Leibler divergence identity applies to probability density functions relative to a common measure when the truthful expected logarithmic loss is finite and the comparison expected value is well defined.^[ref] For sequential forecasts, the existing prequential log score identity explains how logarithmic rewards add over observations.

A strictly increasing function from a closed real interval onto the interval between its endpoint values is continuous. Given a desired output tolerance, choose nearby output values on either side and use their preimages to trap nearby inputs. At the endpoints only one side is needed. Equivalently, a discontinuity of a monotone function creates a gap in its range, which is incompatible with surjectivity onto an interval.

With one A and one B observation per stratum and no ties, the stratified log-rank statistic receives a nonzero contribution only when the earlier observation is a failure while both remain at risk. It is $+1/2$ if A fails first and $-1/2$ if B fails first. After the first observation, only one member remains and every further observed-minus-expected contribution is zero. Under an exchangeable null with independent censoring, conditional on being informative the two signs have equal probability, with mean zero and variance $1/4$. Thus standardization is the normal approximation to a sign test on informative pairs. Failure-time distances do not affect this statistic.

For a finite stopping time $T$, use the shifted filtration ${\mathcal{G}}_s={\mathcal{F}}_{T+s}$. An martingale with a bounded $L^2$ norm shifts to the martingale $M_{T+s}-M_T$; a continuous bounded adapted integrand shifts to a predictable integrand. Their quadratic variation and integration intervals shift by subtracting the value at $T$. Both integrals have identical left-endpoint sums, and the Itô isometry passes the equality to the limit. The starting jump at $T$ is excluded.

Two scalar SDEs driven by the same Brownian motion, with the same globally Lipschitz diffusion coefficient and ordered globally Lipschitz drifts, preserve a strict initial order at every finite time. If $Z=Y-X$ starts positive, apply a reciprocal barrier to the first hit of zero. A Gronwall bound on $(Z+ϵ{)}^{-1}$ makes the hitting probability at most $ϵe^{(L+K^2)t}/(Z_0+ϵ)$, which tends to zero. Common noise and the Lipschitz bound on the diffusion difference are essential to this argument.

In Laplace method for the Gamma function, scaling $t=x(1+s/\sqrt{x})$ gives a standard Gaussian integral. The first even correction polynomial is $s^2-7s^4/12+s^6/18$. Its expectation under a unit-variance centered Gaussian is $1-7\cdot 3/12+15/18=1/12$. The amplitude expansion contributes to this coefficient as well as the phase expansion; expanding only the latter gives the wrong Stirling formula correction.

On the unit sphere, use $\zeta =(x+iy)/(1-z)$ away from the north pole and $\eta =(x-iy)/(1+z)$ away from the south pole. Their overlap relation $\eta =1/\zeta$ is a holomorphic map, giving a holomorphic atlas for the Riemann sphere. The opposite sign of $iy$ in the second manifold chart ensures a holomorphic rather than antiholomorphic transition.

Writing the sphere coordinates as $(X,Y,Z)$ avoids confusing their height with the complex coordinate $z$. The two manifold charts cover the whole unit sphere:On their overlap, $X^2+Y^2=1-Z^2$ givesThe coordinate transformation and its inverse are holomorphic, so these manifold charts make the sphere a Riemann surface. The sign in $X-iY$ is essential. Geometric projection from the south pole with the same planar axis labels would give $v=(X+iY)/(1+Z)=1/\overline{z}$. The complex conjugation in that transition prevents it from being holomorphic. For $z=i$, the correct second coordinate is $w=-i$: positive $Y$ corresponds to negative imaginary part in this manifold chart.

For a Steiner chain between concentric circles, tangencies of neighbouring circles are midpoints of their centres and lie on the circle of radius $\sqrt{R\rho }$. Under a Möbius transformation, their locus is a generalized circle, which may be a line. An ordinary-circle conclusion requires the pole of the inverse transformation to avoid the original locus.

If a simple Steiner chain closes after $n\ge 3$ circles, rotating the starting circle between concentric boundaries gives another closed $n$-circle chain. Applying a Möbius transformation proves the corresponding statement for arbitrary boundary circles. Construction must stay in the same intervening family and proceed without reversing the chosen direction. Merely choosing arbitrary circles tangent to the preceding circle permits backtracking and does not assert closure.

For any nonzero limit ordinal $\delta$, a club set is an unbounded subset closed under limit points below $\delta$. A subset is stationary if it meets every such club set. Statements concerning the completeness of the club filter or Fodor lemma require regular uncountable height; they should not be transferred to arbitrary singular or countable limit ordinals without checking their hypotheses.

A subset $X\subseteq \lambda$ is non-reflecting in the indicated sense if it is not stationary below any $\delta <\lambda$ of uncountable cofinality. The subset itself need not be stationary. Nonreflection is inherited by subsets. With this definition every subset of ${\omega }_1$ is non-reflecting, since all smaller ordinals have countable cofinality.

A local martingale deflator only forces deflated asset prices to be local martingales. True expectation pricing requires additional martingale and integrability properties. In a Brownian one-factor market, continuity of the coefficients ensures a pathwise finite market-price-of-risk square integral, but not expectation-one of its stochastic exponential. The reciprocal three-dimensional Bessel strict local martingale gives a counterexample to promoting this local conclusion automatically.

On a finite state space, the Arrow–Debreu state price is the current price of a payoff equal to one in state $s$ and zero elsewhere. It equals the physical state probability times the state-price density in that state. With a riskless asset of deterministic prices $B_0,B_1$, the Arrow state prices sum to $B_0/B_1$. Normalizing them gives the risk-neutral probabilities, while division by the physical probabilities instead recovers the pricing density.

Let $a>0$, $\lambda >0$, $\sigma \ge 0$ and $\alpha a\ge {\sigma }^2/2$. For an Ornstein-Uhlenbeck process $d{X}_t=\sigma d{W}_t-\lambda X_{t}dt$ under a reference measure, the processis strictly positive and a supermartingale. The Itô formula identifies its short rate asFor $\tau =T-t$, conditional Gaussian second moments give the unit zero-coupon bond priceThis is $\mathbb{E}[{\zeta }_T\mid {\mathcal{F}}_t]/{\zeta }_t$. The money-market density has diffusion coefficient $\sigma X_t/(a+X_t^2/2)$, of magnitude at most $\sigma /\sqrt{2a}$. The Novikov condition therefore validates the corresponding pricing measure on finite horizons. The reference OU dynamics are not automatically the risk-neutral dynamics.

For numbered states of a finite-state automaton, $R_{ij}^{(k)}$ is a regular expression describing paths from $i$ to $j$ whose internal states have numbers at most $k$. Splitting a path at its visits to state $k$ gives $R_{ij}^{(k)}=R_{ij}^{(k-1)}+R_{ik}^{(k-1)}(R_{kk}^{(k-1)}{)}^{\ast }{R}_{kj}^{(k-1)}$. At stage zero use individual edge labels and an empty-word option when the endpoints coincide. This proves the automaton-to-expression direction of Kleene theorem.

For an invertible finite-mode Bogoliubov transformation $a=A{a}^{\prime }+B{a}^{\prime }{}^{†}$, the symmetric squeezing matrix is $M=-A^{-1}B$. The canonical commutation relations make its singular values strictly less than one. The normalized state has $|C|=\det (I-M{M}^{†}{)}^{1/4}$. Infinitely many modes additionally require bosonic mode mixing implementability.

If $e=ir$ and $ri=1$, then $r$ is the coequalizer of $(e,1)$: an arrow $h$ with $he=h$ factors uniquely as $(hi)r$. Conversely a coequalizer $q$ makes $e$ factor as $iq=e$; epimorphic cancellation gives $qi=1$. Thus the splitting of an idempotent morphism is equivalent to this particular coequalizer, without assuming arbitrary coequalizers exist.

For distinct interpolation sites and an invertible B-spline collocation matrix, the interpolant is $P=B{A}^{-1}R$, where $R$ samples data and $Ba={\sum }_i{a}_i{N}_i$. Sampling and basis synthesis have operator norms at most one in the supremum norm. For the reverse inequality, realize the signs of a maximal absolute row sum of $A^{-1}$ as values of a continuous function of norm one, and apply uniform-norm stability of a B-spline basis. This yields both displayed bounds.

In a real-generator convention, $M^{\rho \sigma }=\frac{1}{4}[{\gamma }^{\rho },{\gamma }^{\sigma }]$ obeys $[M^{\rho \sigma },{\gamma }^{\mu }]=g^{\sigma \mu }{\gamma }^{\rho }-g^{\rho \mu }{\gamma }^{\sigma }$. Applying this identity to the commutator of two gamma matrices proves the Lorentz algebra relations. With the matching vector generators and real antisymmetric parameters, $S=\exp (\frac{1}{2}{\Omega }_{\rho \sigma }{M}^{\rho \sigma })$ obeys $S^{-1}{\gamma }^{\mu }S={\Lambda }^{\mu }{}_{\nu }{\gamma }^{\nu }$. A convention using Hermitian Lorentz generators instead must consistently insert factors of $i$.