Adam bias correction compensates for the initial zero values in Adam's exponentially weighted gradient histories. The basic algorithm usesThese factors normalize the coefficient sums of the two running histories.^[ref]

For a scalar coordinate, suppose $m_0=0$ andExpanding the recurrence relation givesIf every $g_s$ has the same expectation $\mu$, linearity of expectation and the geometric series yieldThus ${\hat{m}}_t$ is an unbiased estimator of $\mu$ under these assumptions. Correlation between different times is allowed by this argument. Applying the same calculation to squared gradients with constant raw second moment $\nu =\mathbb{E}[g_s^2]$ gives $\mathbb{E}[v_t]=(1-{\beta }_2^t)\nu$ and $\mathbb{E}[{\hat{v}}_t]=\nu$.

For a constant supplied gradient of 2, choose ${\beta }_1=0.9$ and ${\beta }_2=0.99$. The first uncorrected histories are $m_1=0.2$ and $v_1=0.04$, while their corrected values are 2 and 4. The normalization removes the missing coefficient mass caused by initialization.

It does not make a changing history equal to the latest gradient. For an original deterministic example, let ${\beta }_1=1/2$, $g_1=1$ and $g_2=3$. Then $m_2=7/4$, sowhich differs from $g_2=3$. More generally, changing moment expectations produce a weighted average of past expectations. State the moment assumptions before interpreting “bias-corrected” as unbiased for a present distribution. The nonlinear parameter update built from the corrected histories is a separate object from either moment estimator.

The diagonal AdaGrad accumulator stores a cumulative squared-gradient total for each coordinate:These quantities form part of AdaGrad's optimizer state. An unconstrained diagonal update uses them through a coordinate scale such as $1/(\sqrt{s_{t,i}}+ϵ)$ with a declared positive stabilizer $ϵ$. The accumulator itself and its inverse-square-root transformation are distinct objects.^[ref]

The relation to the full gradient history is exact. Form the sum of outer productsThen $(G_t{)}_{ii}=s_{t,i}$. Retaining only this diagonal is cheaper than retaining every entry of $G_t$, but discards cross-coordinate information.

For an original two-step example, compare historiesBoth give $s_2=(2,2{)}^T$. Their full matrices differ:The diagonal state cannot distinguish those joint gradient patterns. If both are next supplied with $g_3=(1,0{)}^T$, both accumulators become $(3,2{)}^T$ and produce the same coordinate scaling for that update.

Every entry is nonnegative and nondecreasing, since each added square is nonnegative. Alternating gradient signs therefore cannot cancel their contribution to this state. For example, scalar gradients 3 and −3 give accumulated value 18 rather than 0. A saved parameter vector does not reconstruct that history: resuming the same AdaGrad recurrence requires the corresponding accumulator and settings in the optimization checkpoint.

Active power, also called real power, is the time average of instantaneous power transferred through a terminal pair. For periodic electric potential difference $v(t)$ and electric current $i(t)$ of common period $T$, the passive sign convention givesThe net energy entering during each period is $PT$. Positive active power denotes net absorption under this convention; it can represent useful conversion to other forms of energy as well as heating.

For sinusoids of RMS magnitudes $V$ and $I$ with relative phase $\phi$, $P=VI\cos \phi$. For a pure resistor with constant resistance $R$, the stronger waveform-independent statement follows from $v(t)=Ri(t)$:This is why the root mean square current is the constant-current value giving the same heating in that resistor.^[ref]

An original harmonic example shows why RMS values alone do not determine active power. Letin volts and amperes. Orthogonality over a period gives $I_{\mathrm{rms}}=\sqrt{3^2+4^2}=5\mathrm{A}$, but the third harmonic averages to zero when multiplied by the fundamental potential difference. Consequently $P=100(3)=300\mathrm{W}$, whereas apparent power is $100(5)=500\mathrm{VA}$. The power factor is0.6 even though the fundamental current is exactly in phase with the fundamental potential difference.

More generally, Fourier analysis of real periodic waveforms with RMS harmonic magnitudes giveswhen the expansion and average are justified. Cross terms of different harmonic orders vanish. A single fundamental-frequency complex electrical power therefore need not capture every harmonic contribution to the total average.

For an ideal inductor or capacitor in periodic steady operation, the stored energy returns to its starting value each cycle, so its net active power is zero despite nonzero exchange during the cycle. Real components can also have losses. The primary MIT treatment explicitly separates resistor dissipation from periodic inductive and capacitive energy exchange.^[ref]

Complex electrical power represents the active and reactive parts of sinusoidal electrical energy transfer in one complex number. Using phasors $V$ and $I$ whose magnitudes are root mean square values, with the passive sign convention, defineThe star denotes a complex conjugate. The real part is active power; the imaginary part is reactive power. The magnitude $|S|=|V||I|=\sqrt{P^2+Q^2}$ is apparent power. A source using peak-amplitude phasors instead has the factor $1/2$: $S=V_{\mathrm{peak}}{I}_{\mathrm{peak}}^{\ast }/2$.^[ref]

To see why the conjugate matters, write $V=|V|e^{j\alpha }$ and $I=|I|e^{j\beta }$. Then $S=|V||I|e^{j(\alpha -\beta )}$ depends on the relative phase. Changing the common time reference multiplies both phasors by $e^{j\theta }$, leaving $V{I}^{\ast }$ unchanged. The product $VI$ would instead acquire $e^{2j\theta }$ and would not represent a time-reference-independent average-power quantity.

For an original example, choose $V=100\mathrm{V}$ and $I=(4-j3)\mathrm{A}$. ThenThus active power is400W, reactive power is300 reactive volt-amperes and power factor is0.8. The corresponding electrical impedance is $V/I=16+j12\Omega$. With RMS current magnitude5A, the resistive term absorbs $16(5^2)=400\mathrm{W}$, agreeing with the real part.

Under the stated sign convention and $e^{j\omega t}$ time convention, an inductive reactance gives positive $Q$ and a capacitive reactance negative $Q$. Negative $Q$ does not mean a negative dissipated-energy average. For example, adding an ideal parallel capacitive branch contributing $-j300\mathrm{VA}$ cancels this example's reactive part, leaving terminal complex power400VA and RMS source current4A at100V, while the original resistive dissipation stays400W.

This one-frequency representation must not be applied indiscriminately to a distorted waveform. Different harmonics need their own phase relationships; the definition of active power remains the time average of the actual instantaneous power.

The power factor is the ratio of active power to the product of RMS terminal electric potential difference and electric current:when the denominator is positive. It measures how effectively those terminal magnitudes support net average transfer. With signed absorbed power it can be negative for net generation; for an absorbing load the usual range is zero to one. The Cauchy-Schwarz inequality applied to the time average of $v(t)i(t)$ proves $|P|\le V_{\mathrm{rms}}{I}_{\mathrm{rms}}$.

For one-frequency sinusoids, $P=V_{\mathrm{rms}}{I}_{\mathrm{rms}}\cos \phi$, so the power factor is $\cos \phi$. The phase-only quantity for the fundamental is also called displacement power factor. The primary treatments make the RMS convention explicit.^[ref][ref]

A distorted current can give another result. For purely sinusoidal supply potential difference and a zero-mean current with fundamental RMS magnitude $I_1$,Only the in-phase fundamental contributes to the average product with that supply waveform. For an original example with fundamental3A in phase and third-harmonic RMS magnitude4A, total RMS current is5A. The displacement power factor is one but total power factor is0.6. Changing the fundamental phase alone would not remove this harmonic-current contribution.

At fixed absorbed active power and supply RMS potential difference, $I_{\mathrm{rms}}=P/(V_{\mathrm{rms}}\mathrm{pf})$. For an original900W,230V example, changing the factor from0.6 to0.9 changes source current from about6.52 to4.35A. In the same upstream resistance, the corresponding $I^{2}R$ loss becomes $(0.6/0.9{)}^2=4/9$ of its former value. This is a stipulated circuit calculation, not an assertion about a particular tariff or installation.

Power factor correction addresses the mechanism causing the low ratio. An ideal capacitive branch can offset inductive reactive power at one frequency; it does not automatically correct an arbitrary distorted current. State the waveform conditions and whether the reported number is total or displacement factor when comparing measurements.

The harmonic active power sum expresses average electrical power as the sum of contributions from matching components of periodic terminal electric potential difference and electric current. With a common fundamental frequency and root mean square magnitudes for nonzero harmonic components, writeHere $V_0,I_0$ are the signed DC components. The passive sign convention makes the average product absorbed active power. Averaging over a complete fundamental period givesFor finite sums this follows from the cosine-product identity: products of different integer harmonic components average to zero, and matching components give half the cosine of their relative phase. For square-integrable periodic waveforms, the corresponding Fourier series identity follows by continuity of the inner product in the mean-square norm. The orthogonality is a property of the averaging interval and basis, not an assumption that the physical sources are independent.

For an original example, set $V_0=10\mathrm{V}$, $I_0=2\mathrm{A}$, $V_1=100\mathrm{V}$, $I_1=3\mathrm{A}$, and fundamental relative phase $\pi /3$. Add third-harmonic magnitudes $V_3=20\mathrm{V}$, $I_3=1\mathrm{A}$ with relative phase $\pi$. ThenThe third-harmonic contribution is $-20\mathrm{W}$, even though the total absorbed active power is positive. The root mean square values satisfy $V_{\mathrm{rms}}^2=10500{\mathrm{V}}^2$ and $I_{\mathrm{rms}}^2=14{\mathrm{A}}^2$, so apparent power is about $383.41\mathrm{VA}$ and total power factor about $0.3912$.

For a purely sinusoidal supply with no DC component, only $V_1$ survives, giving $P=V_1{I}_1\cos ({\alpha }_1-{\varphi }_1)$ even if the electric current is distorted. Current in other harmonic components can still increase conductor heating. The usual one-frequency complex electrical power construction is derived in the primary MIT treatment; the sum above applies that average-product reasoning separately to orthogonal components.^[ref]

A proposed measurement records the two waveforms synchronously and compares their directly averaged product with the resolved sum. A non-integer-period window, unresolved harmonic components, channel timing error or changing waveform can spoil the simple finite-record equality. The formula assumes a shared periodic model; multiplying separately measured RMS magnitudes gives apparent power instead.

Over an algebraically closed field $k$, an ample line bundle $L$ on an abelian variety $A$ defines a homomorphismwhere $A^{\vee }$ is the dual abelian variety and $t_x$ is translation by $x$. This isogeny is a polarization of an abelian variety. Over a general field, a polarization is an isogeny that has this form after extending to an algebraic closure; an inducing ample line bundle need not descend to the original field.^[ref]

A principal polarization is an isomorphism $A\to A^{\vee }$. An elliptic curve with its degree-one origin divisor gives a basic example. A polarization supplies positivity data relating the geometry of line bundles to the algebra of isogenies; twisting $L$ by a degree-zero line bundle does not change ${\lambda }_L$.

2D IR response-factor transferability means that a response ratio characterized under one set of conditions remains applicable to another measurement. It is an assumption to test when using a 2D IR response-factor ratio to infer populations from spectral amplitudes. A common detector gain can cancel from a ratio, while different gains or optical responses for its two components do not.

In the positive-component measurement model, writeThe shared survival factor $P$ cancels. If conditions change the responses to ${\alpha }^{\prime }=g_D\alpha$ and ${\beta }^{\prime }=g_C\beta$, then ${\kappa }^{\prime }=(g_C/g_D)\kappa$. Equality of $g_C$ and $g_D$ is a substantive requirement, not a consequence of both signals coming from the same instrument.

An original error calculation supposes a reference established ${\kappa }_0=2$, but the unknown measurement actually has ${\kappa }_1=3$. For true population fraction $q=0.20$, its observed amplitude ratio is $C/D=3(0.2)/0.8=0.75$. Using the old response ratio givesan overestimate even with noiseless detection. More generally, with $\lambda ={\kappa }_1/{\kappa }_0$,A small response-ratio mismatch $\delta \ln \kappa$ therefore gives $\hat{q}-q\simeq q(1-q)\delta \ln \kappa$.

The optical reason to check transfer is visible in the primary response theory: different pathways contain different transition-dipole products, and their amplitudes include relaxation factors. Pulse bandwidth, polarization and the selected 2D IR waiting time help determine which response is recorded.^[ref] This does not establish a universal two-population model for every 2D IR cross peak.

A 2D IR response-factor validation mixture supplies independently known composition under the new conditions. Test whether the old response ratio predicts its amplitudes, and repeat over an appropriate composition range if matrix effects or nonlinearity are plausible. Reusing an unknown's fitted fraction to validate the factor that produced it would be circular.

Retain the amplitude definition and phase convention as well as the instrument settings. The phase-controlled peptide experiment measures phases and amplitudes, so a signed or complex component need not obey the positive-component equations above.^[ref] Successful transfer concerns a specified model and observable, not every intensity visible in a spectrum.

A 2D IR slope-axis convention specifies which frequency coordinate is held fixed when taking spectral slices, which coordinate supplies each slice maximum, and which coordinate is plotted vertically. It also specifies whether the reported 2D IR center-line slope is that geometric slope or its inverse. These choices are needed before comparing slopes from two-dimensional infrared spectroscopy experiments or software.

Let $x$ be an excitation-frequency offset and $y$ a detection-frequency offset. An original positive bivariate normal distribution envelope with widths ${\sigma }_x,{\sigma }_y$ and correlation coefficient $\rho$ has conditional maximaThe first relation maximizes over $y$ at fixed $x$; the second maximizes over $x$ at fixed $y$. They trace different lines unless the distribution is degenerate. Changing the slicing direction is not the same operation as taking the reciprocal of an already fitted line's slope.

For $\rho =0.6$, ${\sigma }_x=10$ and ${\sigma }_y=20$ in matching frequency units, the first slope is 1.2. The second conditional slope $dx/dy$ is 0.3. Plotting that second set of maxima with $y$ vertical gives geometric slope $dy/dx=1/0.3=10/3$; its inverse is 0.3, not 1.2. Both sets came from the same stipulated band.

The 2007 primary method takes cuts parallel to the initial-frequency axis and defines its reported quantity through an inverse-slope convention. Its relationship with a vibrational frequency-frequency correlation function is established within the paper's response approximations.^[ref] A generic graphing routine's slope label does not identify that observable.

Units can change a geometric slope without changing the underlying correlation coefficient. Under coordinate scalings $x^{\prime }=ax$ and $y^{\prime }=by$, a fitted slope becomes $m^{\prime }=(b/a)m$. Expressing only the vertical axis in GHz instead of THz multiplies its numerical slope by 1000. Using the same conversion on both axes leaves the slope unchanged. Reciprocal wavelength coordinates require their own transformation rather than a constant rescaling over an arbitrary bandwidth.

Report axis labels, units, slice direction, fit range and inversion convention with the result. At $\rho =0$ in the ideal envelope, the maxima at fixed $y$ lie on $x=0$: that line is vertical even though its inverse slope tends to zero. A failed finite geometric fit near this limit need not mean molecular frequency memory increased. Noise, overlapping signed bands and finite fitting ranges require separate checks.

The tangent modulus is the local slope of a specified uniaxial stress-strain curve:It converts a small additional strain near ${\epsilon }_0$ into the corresponding first-order change of stress. For nonlinear elastic deformation, this slope can vary with the operating point.^[ref] In uniaxial linear elasticity, $\sigma =E\epsilon$, so the tangent modulus equals the constant Young's modulus.^[ref]

For an original nonlinear branch, stipulateThen $E_{\tan }=E_0+2B\epsilon$. At ${\epsilon }_0=0.02$, the stress is $80\mathrm{MPa}$ and the tangent modulus is $6\mathrm{GPa}$. The secant modulus from the stipulated unloaded origin is instead $\sigma /{\epsilon }_0=4\mathrm{GPa}$. The two slopes answer different questions: the secant compares total stress and strain, while the tangent predicts a small further increment.

For $\Delta \epsilon =0.001$, the tangent predicts $\Delta \sigma \simeq 6\mathrm{MPa}$. The exact change in the chosen polynomial is $6.1\mathrm{MPa}$ because the omitted term is $B(\Delta \epsilon {)}^2=0.1\mathrm{MPa}$. This is a linear approximation whose error decreases quadratically with the increment, not a claim that the nonlinear material obeys Hooke's law globally.

To estimate a tangent modulus experimentally, specify the stress and strain measures, temperature, loading history and loading rate. For engineering stress $\sigma =F/A_0$ and engineering strain $\epsilon =\Delta L_g/L_g$ from a local extensometer,Fit a local window about the chosen operating point and check how the estimate changes with that window. For an original $L_g=50\mathrm{mm}$, $A_0=10\mathrm{m}{\mathrm{m}}^2$ and local slope $1200\mathrm{N}\mathrm{m}{\mathrm{m}}^{-1}$, this gives $6000\mathrm{N}\mathrm{m}{\mathrm{m}}^{-2}=6\mathrm{GPa}$. Differentiation amplifies measurement noise, and crosshead displacement can include machine compliance. Where plastic deformation or history-dependent response occurs, identify the loading branch: its local slope need not equal the unloading slope or an elastic material constant.

The secant modulus at a specified point of a stress-strain curve is the slope of the line from the declared unloaded origin to that point:It depends on the chosen point when the curve is nonlinear. Mississippi State University's solid-mechanics laboratory material defines it geometrically in this way.^[ref] Specify both the strain measure and the origin when reporting a secant modulus.

For an original branch $\sigma =E_0\epsilon +B{\epsilon }^2$, with $E_0=2\mathrm{GPa}$ and $B=100\mathrm{GPa}$,At ${\epsilon }_{\ast }=0.02$, stress is $80\mathrm{MPa}$, the secant modulus is $4\mathrm{GPa}$ and the tangent modulus is $6\mathrm{GPa}$. The first slope relates accumulated stress and strain; the second predicts a small further increment. In linear elasticity through the unloaded origin they both equal Young's modulus.

The relation is especially clear for a smooth curve with $\sigma (0)=0$. The fundamental theorem of calculus givesThus the secant modulus averages the local tangent slopes over that interval. A chord modulus instead joins two specified states, with slope $({\sigma }_2-{\sigma }_1)/({\epsilon }_2-{\epsilon }_1)$; its first state need not be the unloaded origin.

For a tensile measurement using initial cross-sectional area $A_0$ and local gauge length $L_g$, engineering stress and engineering strain givewhere extension is measured from the appropriate unloaded state. An original $F=800\mathrm{N}$, $A_0=10\mathrm{m}{\mathrm{m}}^2$, $L_g=50\mathrm{mm}$ and $\Delta L_g=1\mathrm{mm}$ yields $E_{\sec }=4000\mathrm{N}\mathrm{m}{\mathrm{m}}^{-2}=4\mathrm{GPa}$. A local extensometer avoids attributing all machine compliance to specimen extension. Retain the loading history, temperature and target stress or strain: a secant from an arbitrary preload and an unloading slope are different measurements, particularly after plastic deformation.

A chord modulus is the slope of the line connecting two specified points on a stress-strain curve:Instron's glossary defines the calculation by selecting a starting strain and an ending strain and taking the slope between their curve points.^[ref] Report the endpoints and strain definition with the value. A linear regression through every observation between them is generally a different estimator.

For an original branch $\sigma =E_0\epsilon +B{\epsilon }^2$, subtraction and factorization giveWith $E_0=2\mathrm{GPa}$, $B=100\mathrm{GPa}$, ${\epsilon }_1=0.01$ and ${\epsilon }_2=0.03$, the endpoint stresses are $30$ and $150\mathrm{MPa}$, so $E_{\mathrm{chord}}=6\mathrm{GPa}$. The secant modulus from the unloaded origin to the second point is instead $5\mathrm{GPa}$. For this quadratic branch, the chord happens to equal the tangent modulus at the midpoint strain; this midpoint equality does not hold for every nonlinear curve.

More generally, the fundamental theorem of calculus gives, for a smooth branch,Thus a chord samples the response across a finite interval. Shrinking that interval can approach a tangent modulus, but measurement noise becomes more consequential when divided by a small strain difference.

For engineering stress and local engineering strain measured with an extensometer,An original specimen with $A_0=10\mathrm{m}{\mathrm{m}}^2$ and gauge length $L_g=50\mathrm{mm}$ gives $6\mathrm{GPa}$ when force changes from $300$ to $1500\mathrm{N}$ and local extension changes from $0.5$ to $1.5\mathrm{mm}$. A constant readout offset cancels in each difference when those same physical states are used; scale errors and machine compliance do not generally cancel. Specify temperature, loading direction and history, since a chord across different branches need not describe one material response.

A flexural modulus is a stiffness inferred from a bending experiment using a specified stress–strain or force–deflection model. It has units of pressure. In a homogeneous, isotropic, linearly elastic slender beam with negligible shear and fixture deformation, it agrees with the Young's modulus entering $M=EI\kappa$. It is not the flexural strength or the load at fracture.

For a simply supported uniform beam in a symmetric four-point flexural test, let outer span be $L$, loading contacts lie at $a$ and $L-a$, and $P$ be their total force. Integration of the small-deflection Euler-Bernoulli beam equation gives midpoint deflection magnitudeEquivalently, with measured linear slope $S=dP/d\delta$,The formula uses midpoint displacement relative to the supports, not automatically the testing machine's crosshead displacement. The underlying moment-curvature law is the beam relation derived by Roylance; actual test conversion additionally requires the fixture spans and specimen geometry.^[ref][ref]

For an original rectangular example, choose $L=120\mathrm{mm}$, $a=30\mathrm{mm}$, $b=10\mathrm{mm}$ and $h=4\mathrm{mm}$, so $I=53.333\mathrm{m}{\mathrm{m}}^4$. A corrected midpoint slope $S=100\mathrm{N}\mathrm{m}{\mathrm{m}}^{-1}$ gives $E_{\mathrm{flex}}\approx 46.4\mathrm{GPa}$. This is an inferred model value from stipulated inputs, not a tabulated property of a named material.

If measured displacement includes an extra compliance $C_{m}P$, the measured compliance is $d{\delta }_{\mathrm{meas}}/dP=C_{\mathrm{beam}}+C_m$. Ignoring positive fixture compliance reduces the apparent inferred modulus. Transverse shear, anisotropy, nonlinear response and time-dependent material behavior can also make flexural and tensile measurements differ. A reported value should therefore identify geometry, slope convention, displacement measurement and loading conditions.

A microquasar is an accreting stellar binary whose compact object launches relativistic jets, providing a small-scale analogue of a jet-producing active galactic nucleus. The name refers to the analogy between accretion and outflow, not to a miniature galaxy. The accretor may be a black hole or a neutron star.

A revealing observation is a sequence of high-resolution radio images, showing emitting knots moving away from the compact central source. In GRS 1915+105, such observations demonstrated strikingly rapid apparent motions. If a knot moves at speed $\beta c$ at angle $\theta$ toward the observer, light-travel delays giveFor the illustrative values $\beta =0.9$ and $\theta =30^{\circ }$, this is approximately $2.04$: apparent superluminal motion despite a locally subluminal speed. The receding jet has a different arrival-time factor. Comparing both sides and the source distance therefore constrains geometry, rather than measuring a speed directly from one image.^[ref]

A Wyckoff letter is the conventional alphabetic label identifying a Wyckoff position within the tables for a space group. It is normally written beside the crystallographic site multiplicity, as in $1a$. Letters begin with $a$ for a position with maximal site-symmetry group order and proceed toward the general position. The letter is a label, not a coordinate or a chemical-species symbol.^[ref]

For example, the space group $P1$ has only translations. Every point has a trivial site-symmetry group, and the general position is labeled $1a$: one equivalent position per primitive unit cell, with free coordinates $(x,y,z)$.^[ref]

An original model makes the distinction from atom counting clear. Place one species at $(0.1,0.2,0.3)$ and another at $(0.4,0.1,0.7)$ in a generic $P1$ crystal structure. Both sites are described by $1a$, but they define different crystallographic orbits. The notation does not say that the entire structure contains only one atom per cell.

The space group, its setting and coordinate convention must accompany the letter. The same letter in another group need not denote the same symmetry or coordinates. Nor does matching multiplicity by itself identify the Wyckoff position, since distinct positions can have equal multiplicity.

A wringing-film covariance model specifies how uncertain effective contact-layer contributions vary together in a gauge block wringing measurement. Individual contact uncertainties do not determine the uncertainty of a stack unless their dependence is also specified.

Let the vector of contact differences be $\mathbf{\delta }=({\delta }_1,\dots ,{\delta }_n{)}^T$ and its covariance matrix be $\Sigma$. For a correction $\Delta ={\mathbf{c}}^T\mathbf{\delta }$, propagation givesThe coefficients encode the actual geometry and calibration subtraction. They need not all be positive. This is the general correlated-input law applied to wringing-film uncertainty.^[ref]

A simple original model writes ${\delta }_j=b+{ϵ}_j$, where one common contribution $b$ has variance $u_b^2$ and independent contact contributions have variance $u_{ϵ}^2$. ThenFor three equal-sign contacts with $u_b=3\mathrm{nm}$ and $u_{ϵ}=2\mathrm{nm}$, the variance is $93\mathrm{n}{\mathrm{m}}^2$. For a difference of two contacts with coefficients $(1,-1)$, the common contribution cancels and the variance is only $8\mathrm{n}{\mathrm{m}}^2$. These are chosen covariance structures, not universal properties of wrung surfaces.

Calibration and use can share only part of their contact history. If $n$ calibration contacts share effect $b_c$ and $n$ use contacts share effect $b_u$, their common-effect correction is $n(b_u-b_c)$. Its variance isPerfectly shared equal contributions cancel; independent preparation effects do not. Assuming either relation requires evidence about preparation, reference surfaces and measurement history.

Repeated readouts of one unchanged wring mostly probe readout scatter. Rewringing samples a different level of variability; changes of surface preparation or elapsed time may add another. The NIST handbook documents that contact layers depend on conditions and can evolve after thermal equilibration.^[ref] A nested wringing-repeatability study can help separate those levels, but repeatability data alone do not identify every shared calibration contribution.

Retain the covariance assumptions with the result and test whether the proposed matrix is a valid positive semidefinite matrix. A small combined uncertainty obtained by assigning cancelling correlations is meaningful only when those correlations describe the physical comparison.

The central axis of a wrench is the line of reference points about which the resultant torque is parallel to the nonzero resultant force. It converts a general wrench into a force along one line plus a parallel force couple. A nonzero parallel couple cannot be removed merely by moving the force’s application point.

Let $\mathbf{F}\ne 0$ and ${\mathbf{\tau }}_O$ be the force and moment about origin $O$. DefineThen the central axis, expressed relative to $O$, isHere $h$ is the wrench pitch. Since ${\mathbf{r}}_0\cdot \mathbf{F}=0$, ${\mathbf{r}}_0$ is the point on the axis closest to $O$. The primary robotics wrench representation supplies the force/moment translation law used in this construction.^[ref]

For an original derivation, write the moment at a point displaced by $\mathbf{r}$ as ${\mathbf{\tau }}_{\mathbf{r}}={\mathbf{\tau }}_O-\mathbf{r}\times \mathbf{F}$. The triple-product identity gives ${\mathbf{r}}_0\times \mathbf{F}={\mathbf{\tau }}_O-h\mathbf{F}$, henceThe two terms are perpendicular. ThereforeThis proves both the minimum-moment property and the uniqueness of the minimizing line: the minimum magnitude is $|h|\Vert \mathbf{F}\Vert$, attained exactly on $\mathcal{L}$. It also quantifies the moment penalty for a perpendicular displacement from that line.

For a numerical example, take $\mathbf{F}=(10,0,0)\mathrm{N}$ and ${\mathbf{\tau }}_O=(2,3,-4)\mathrm{N}\mathrm{m}$. Then $h=0.20\mathrm{m}$, ${\mathbf{r}}_0=(0,0.40,0.30)\mathrm{m}$, and the axis is parallel to the $x$ direction. The closest-axis distance from the original origin is $0.50\mathrm{m}$. On the axis the moment is $(2,0,0)\mathrm{N}\mathrm{m}$; at a point $0.10\mathrm{m}$ perpendicularly from it, its magnitude is $\sqrt{2^2+1^2}\approx 2.236\mathrm{N}\mathrm{m}$. At $O$, it is $\sqrt{29}\mathrm{N}\mathrm{m}$.

If $h=0$, a single force on the central axis represents the wrench. If $\mathbf{F}=0$ and the moment is nonzero, every origin has the same pure couple and this construction has no unique finite axis. With both zero, no direction is selected at all. Near zero force, the denominator also makes a reconstructed axis sensitive to measurement uncertainty; an enormous calculated offset need not be a well-resolved physical location.

A proposed six-component load-sensor test applies a known force on an offset line and a known parallel couple, then computes the axis from measurements about several reference points. Correctly translated axes should describe the same geometric line even though their coordinate vectors differ. Subtract tare loads and retain axis conventions before comparing the resulting moments; minimum moment does not mean zero moment unless the wrench pitch is zero.

A synset in WordNet is an unordered set of expressions grouped around one specified meaning. It contains a definition and usually examples, while relations connect it to other synsets. This differs from treating every meaning of one spelling as a single record.^[ref]

The provider illustrates synonymy with car and automobile. Their shared concept can be represented once while retaining both expressions. Conversely, a written form with several meanings participates in different synsets.^[ref]

For an original toy design, let one record contain two expressions for a vehicle and another contain an expression for a boat. There are three expression entries but two concept records. Counting expressions and counting meanings consequently answer different questions. The database's grouping is an explicit semantic analysis: it is useful for matching and navigation, but does not prove universal interchangeability in all registers or constructions.

Common-mode support interference in mounting comparisons is an aerodynamic disturbance shared by different mounting arrangements, so that agreement between their corrected loads fails to reveal it. Repeating a wind tunnel measurement with several supports can test differences in support effects without identifying their shared absolute contribution.

Write the support-tare-corrected load for mount $j$ aswhere $F$ is the desired isolated-model force, $b$ a shared wind-tunnel support interaction residual, ${\delta }_j$ a mount-specific residual and ${\epsilon }_j$ measurement noise. Differences eliminate both $F$ and $b$:Even arbitrarily precise agreement does not separate $F$ from $b$. Replacing $(F,b)$ by $(F+c,b-c)$ leaves every predicted observation unchanged. This is shared-bias nonidentifiability in mounting comparisons.

For an original example, corrected mount results 21.9, 22.0 and 22.1 N are compatible with isolated-model force 20.0 N, common residual 2.0 N and mount-specific residuals $-0.1,0,0.1$ N. Averaging reduces their visible spread but does not remove the common 2 N. A standard error derived only from replicate spread would omit that contribution.

An independent bound adds information. If a reference experiment justifies $1.5\le b\le 2.5$ N and other residuals are negligible, an exact mean 22 N implies $19.5\le F\le 20.5$ N. Without the independent bound, agreement alone supplies no such interval. The reference must constrain the common disturbance, rather than repeat the same unvalidated subtraction.

Real support interference altered both airfoil drag coefficients and their inferred thickness dependence in the 1939 variable-density-tunnel investigation.^[ref] Multiple mounting arrangements are useful diagnostics, but their concordance establishes only the aspects to which the comparison is sensitive.