Let index either of two candidate pathways and suppose for a measured symptom . For any mixing probability , marginalization gives . An independent sample with positive observations out of therefore has likelihood , independent of . The symptom alone does not identify the mixture proportion or pathway. This is an identifiability counterexample, not an empirical claim that real clinical pathways necessarily have identical symptom distributions. Additional indicators or interventions may distinguish them.
Synaptic release probability is the probability of release at a specified site during a specified opportunity, such as one presynaptic action potential. It depends on the site's state and conditions; it is not the probability that the receiving neuron fires. In the example under synaptic vesicle, refers to one available site, whereas refers to any of four independent sites releasing.
Synaptic vesicle 2026-10-14
A synaptic vesicle is a small membrane-bounded compartment in a presynaptic terminal that stores neurotransmitter for release. Fusion with the cell membrane exposes its contents to the synaptic cleft; the packet supplied by one synaptic vesicle is a quantum of release. Quantal neurotransmitter release refers to these discrete packets, rather than the release of one molecule at a time. The postsynaptic response to a packet can vary with its contents, receptors and recording conditions. See Purves and colleagues: vesicles as the source of transmitter quanta.
The useful cycle is filling, preparation for fusion, calcium-triggered exocytosis, and recovery of membrane and proteins through endocytosis for reuse. Different recovery routes occur; recycling need not preserve a particular synaptic vesicle as an identifiable container throughout. Limited availability of prepared synaptic vesicles can reduce release during repeated stimulation, while changes in synaptic release probability can also change successive responses. See Basic Neurochemistry: secretion machinery and vesicle reuse and 3.1 Synapses.
For an illustrative calculation, assume four release sites, each with one prepared synaptic vesicle, each releasing with synaptic release probability during one presynaptic action potential. Assume independent Bernoulli trials and at most one release per site. The released packet count has a binomial distribution:
The probability of no release is ; at least one release has probability . An expected value of one packet does not mean that every action potential releases one packet. If each packet contributes a stipulated conductance increment and increments add, the expected value of that increment is , with zero increment on about of trials. These assumptions are a teaching model: shared calcium fluctuations, unequal sites, depletion and nonlinear receptor responses can invalidate them.
Take independent copies of the simple random walk on the integer line, each stepping left or right with equal probability at every integer time. Their product Markov chain moves all coordinates at once; each of its diagonal increments has probability . At odd times a return to the origin is impossible. At time , independence gives the displayed return probability, whose asymptotic equivalence follows from Stirling's approximation. The recurrence criterion by return probabilities and the p-series therefore imply recurrence for and a transient Markov chain for .
For the chain from the origin visits only the sublattice whose coordinates all have the same parity. This is one communicating class; restricting to it gives an irreducible Markov chain. The period of a state in a Markov chain is two here. The conclusion applies to that class, without incorrectly claiming Markov chain irreducibility on all of . This walk moves diagonally; the usual axis-step simple random walk has different transition probabilities, even though the same dimensional recurrence threshold holds.
Tagged-label Rayleigh quotient for adjacent transpositions Created 2026-10-06 Updated 2026-11-05
For the random adjacent transposition shuffle on with , the position of one label is uniform on in equilibrium, so . Each adjacent transposition moves the label with Probability , giving . The Poincare inequality for a reversible Markov chain variational formula therefore gives . In particular the continuous uniform distribution on has variance , not .
Tail event Created 2026-09-24 Updated 2026-10-14
A tail event is an event , where is the tail sigma-algebra of the chosen sequence of random variables. It concerns information available arbitrarily far along the sequence. In the measurable coordinate models described in finite-prefix invariance of tail events, any finite edit preserves whether it occurs.
This is stronger than merely needing a long waiting time. On , with denoting heads, let , taking if there is no head. The event
is not a tail event: changing only from zero to one changes membership. Under fair independent Bernoulli trials its probability is . Even the event , of probability one in this model, is not a tail event as a literal set: the all-zero sequence and the sequence with just one initial head differ in one coordinate. A probability of zero or one does not characterize tail events.
Tail measurability of limits of sample averages Created 2026-10-05 Updated 2026-10-14
Let be a sequence of finite real random variables, and write for its sample mean. For fixed and , put . Then, at each outcome,
Consequently . The right-hand limit superior is a measurable function of for every , so the extended-real random variable is measurable with respect to the tail sigma-algebra. The same holds for the limit inferior. Thus the event of convergence to a finite limit of a sequence, and the event for any specified real , are tail events.
For a worked deterministic example, let alternate , and replace its first three entries by . For every the two sample means differ by exactly , so both converge to . Finite edits can change small-sample answers while preserving the infinite limit of a sequence.
Under independence, Kolmogorov's zero–one law makes each specified convergence event have probability zero or one. It does not select which value occurs. If a finite random variable measurable with respect to the tail sigma-algebra exists, the events for rational each have probability zero or one; continuity of probability and their nesting force to be constant almost surely. For fair independent Bernoulli trials, the strong law of large numbers establishes almost surely; the zero–one conclusion alone proves neither convergence nor the value .
Tail probability Created 2026-10-03 Updated 2026-10-14
For a real random variable , an upper-tail probability has the form and a lower-tail probability has the form . Each is the probability of an event lying beyond a chosen threshold in one direction. A tail probability need not be small: the threshold determines how much of the probability distribution it includes. A concentration inequality bounds such probabilities, often for deviations from an expected value.
Endpoint conventions matter. Write for the cumulative distribution function and . The complement rule of probability gives four distinct formulas:
The strict upper-tail convention is the survival function . Its value differs from the inclusive upper tail probability by exactly the probability mass at the threshold:
A singleton with positive probability mass is an atom of a measure. Thus excludes an atom of a measure at , whereas includes it. These endpoint identities follow directly from the cumulative distribution function properties in Kyle Siegrist, Distribution and Quantile Functions: endpoint probabilities and right tails.
As the threshold increases, both upper tail probabilities are nonincreasing functions. The strict survival function is a right-continuous function; the inclusive upper tail probability takes the upper endpoint value at a downward jump, equal to the left limit of the survival function. Both tend to zero as for a real random variable, but the speed of that decay differs across probability distributions. The cumulative distribution function properties and their proof from continuity of probability measures are given in MIT 6.436J, Lecture 4: random variables and cumulative distribution function properties.
For a fair six-sided die , the discrete uniform distribution assigns probability mass to each of . At threshold four,
Between four and five, both upper tail probabilities equal . At the largest possible outcome the distinction is especially visible: , but . The diagram shows exact probabilities for this mathematical model; its dots specify which endpoints belong to each step function.
Figure 1.
Exact fair-die probabilities: including or excluding a threshold changes the tail by one atom
. Details.
If a probability distribution has a probability density function , then every singleton has zero probability mass and
For an exponential distribution of rate two, for and zero otherwise, so for and for . Hence . This tail probability is an integral of the probability density function, not its height . A mixed probability distribution retains endpoint jumps even with a smooth component: give zero probability mass and choose this exponential distribution with probability . Then , , and for .
A two-sided tail probability can measure the absolute value of displacement from a specified center . For , the two extreme events are disjoint events, so
For strict displacement , replace the last expression by . The condition prevents overlap; at adding inclusive tails double-counts any probability mass at . For a uniform distribution on , take and : the two tail probabilities are each and their sum is . Equal tails follow from this symmetric probability distribution, not from the phrase two-sided alone; for a uniform distribution on with , the two tail probabilities are zero and .
The tail-sum formula for expectation turns a whole curve of tail probabilities into an expected value. For any nonnegative random variable , without assuming a probability density function or a finite expected value,
For a proof outline, write pointwise and exchange the expected value and Lebesgue integral using Tonelli theorem; is an indicator function. The inclusive version has the same Lebesgue integral: positive probability masses occur at only a countable set of thresholds, which has Lebesgue measure zero. For a nonnegative integer-valued random variable , the tail-sum formula instead gives . The integer hypothesis is essential for these sums. The integral identity, its indicator function proof, and the integer version appear in Kyle Siegrist, Additional Properties of Expected Value: integration and summation of right tails.
For the fair die, this tail-sum formula yields , equal to the direct weighted arithmetic mean . For the rate-two exponential distribution, the area under the survival function is , its expected value. The first panel below shades an integral of the probability density function for a single threshold; the second shades an integral of the survival function across all nonnegative thresholds. These are analytical illustrations, with the displayed curves truncated at three; their exact total areas use an infinite integration range.
Figure 2.
Exponential model: one tail probability is density area, while expectation is total survival area
. Details.
Nonnegativity is essential to interpreting the upper tail probability area as : for a general real random variable that area equals the expected value of its positive part. An integrable random variable satisfies , subtracting the expected value of its negative part; infinite terms cannot be canceled. Even when is finite almost surely, its expected value can be infinite. For the shape-one Pareto distribution on , on and for , so the area is . The finite versus infinite expected value distinction is illustrated in Kyle Siegrist, Definitions and Basic Properties: exponential and Pareto expected values.
Why are tail probabilities useful? They answer threshold questions directly, and a concentration inequality can bound them without knowing the full probability distribution. For nonnegative with finite expected value, Markov inequality gives for : a large threshold relative to the expected value forces rarity. The quantile function reverses the question by choosing a threshold for a prescribed cumulative level. If with , then ; an atom of a measure may prevent equality. For the die, , with strict upper tail probability zero and inclusive upper tail probability .
A tail probability describes rarity under a specified probability distribution; it is not automatically a conditional probability after extra information. For with ,
because the threshold events are nested. Thus the die has but . Conditioning changes the sample space and renormalizes its probability measure; the same extreme outcomes can become common in that restricted population.
Tail sigma-algebra Created 2026-09-24 Updated 2026-10-14
For a sequence of random variables on a probability space , define
The decreasing sigma-algebras are generated by the listed random variables. Their intersection consists of events expressible after discarding every finite prefix. No independence is required.[ref]
The point is to distinguish a long observation record from a property of an infinite sequence. The event of seeing infinitely many heads and the event that sample means converge survive any finite change, while an exact first-head time does not. Kolmogorov's zero–one law adds the striking conclusion that under independence each tail event has probability zero or one; the tail sigma-algebra itself can still contain many nonempty sets.
Tail-sum formula Created 2026-09-24 Updated 2026-10-18
The tail-sum formula expresses an expected value as the area under a survival function. For every nonnegative random variable ,
This Lebesgue integral and the expected value may both be ; a probability density function is unnecessary. For a random variable taking values in the nonnegative integers, the corresponding sum is
The events and coincide because takes integer values. Thus the strict sum starts at zero, while the inclusive sum starts at one.[ref]
For an original example, let the probability mass function of be
Its expected value computed directly is . The strict tail probabilities at thresholds are : their sum is again . There is no probability mass at two, yet the event still has probability . The survival function has heights on the successive intervals ; each interval has length one, so its integral is the same sum.
Figure 1.
The same expected value from probability masses and survival columns
. The exact constructed probability mass function assigns to . The contribution appears in one survival column, while the contribution appears in three. Filled endpoints give the strict survival function at its jumps; open endpoints show the preceding height. The shaded area is . Details.
The indexing matters numerically. Starting the inclusive sum at zero adds , giving for this example. Starting the strict sum at one omits , giving . For a general nonnegative random variable, replacing with inside the integral is harmless: the difference is , nonzero at only a countable set of thresholds. For each positive integer , at most thresholds can have probability mass at least ; their countable union contains all thresholds with positive probability mass and has Lebesgue measure zero. This reasoning concerns an integral, not equality of the two survival function conventions at every point.
For a continuous original example, take the uniform distribution on the interval . Its survival function is for and zero for . The area is a triangle of base four and height one:
Directly integrating against the probability density function gives the same expected value, . The integral of the probability density function alone is one; it is the survival function whose area is the expected value.
Figure 2.
A uniform density has total probability one and survival area two
. Exact curves for the uniform distribution on . The probability density function has integral one, while the survival function has integral two. Splitting the survival area at gives before the cap and beyond it, the expected values of and the positive part , respectively. Details.
The proof counts layers of each outcome. For a realized value , the indicator function equals one over an interval of length , so
Tonelli theorem permits the interchange because the indicator function is nonnegative; no finite expected value is assumed. Similarly, an outcome contributes one indicator random variable for each of the events that occur:
Again Tonelli theorem applies to the nonnegative sum, or equivalently the monotone convergence theorem applies to its partial sums. This exchanges counting layers within each outcome with averaging across outcomes. The existing integer-valued hypothesis in the tail-sum formula explains what changes when the values are not integers; the tail integral formula for moments weights the layers to obtain higher moments.
For an integrable random variable that can have either sign, apply the nonnegative formula separately to its positive part and negative part :
Integrability makes both integrals finite. More generally the difference defines an extended expected value when at least one is finite; if both are infinite, subtraction does not define an expected value.[ref]
Tail probabilities are often easier to measure or bound than a whole probability density function. For a device's survival time, the survival function records the probability of still operating at each threshold; its integral gives the mean operating time. A finite observation window naturally gives the expected value of a capped nonnegative random variable. For a mathematical bound, for all implies . The tail-sum formula turns control of a whole family of events into control of an average.
Talagrand concentration inequality for certifiable functions Created 2026-09-24 Updated 2026-11-05
For a nonnegative integer-valued certifiable function of independent random variables, suppose changing one coordinate changes by at most and a value has a certificate involving at most coordinates, with . Then for ,
For a high-value input , every low-value input must disagree with its certificate on at least coordinates: replacing its certificate positions by those of would change the low value by less than but force a value at least . Uniform weights on certificate positions therefore give Talagrand convex distance at least from the low-value event. Apply Talagrand's convex distance inequality to that event.
The bound is useful when one of the two events has known nonnegligible probability, for example at a median. A certificate alone supplies no coordinate-change constant ; both hypotheses are used in the proof. The two-threshold concentration for certifiable functions article develops the resulting mean-median comparison.
Talagrand's concentration inequality Created 2026-10-14 Updated 2026-11-05
Talagrand's concentration inequality gives strong concentration in product spaces using convex distance. If has positive product probability, its convex-distance enlargement has Gaussian tails. A Talagrand concentration inequality for certifiable functions follows by bounding this distance using bounded differences and small certificates.
Talagrand's convex distance inequality Created 2026-10-06 Updated 2026-11-05
On a finite product of standard probability spaces, for any measurable event of positive probability,
Consequently . Here is the Talagrand convex distance. The distance-event conclusion can be read with outer probability when necessary.
Tangent of a uniform angle has a Cauchy distribution Created 2026-10-18 Updated 2026-10-20
If and has the continuous uniform distribution on , then has the Cauchy distribution with location zero and scale . The tangent function is strictly increasing on this interval, so its cumulative distribution function is
Differentiating gives the probability density function . The probability of either endpoint is zero, so including endpoints in the uniform interval does not alter this probability distribution. The slow quadratic decay makes the expected value undefined: its positive and negative parts both have infinite integrals. Its second moment is also infinite, so it has no finite standard deviation.
Teleportation fidelity 2026-10-20
For a pure input target and normalized output density operator , the displayed squared quantum fidelity measures how well a teleportation channel reproduces that target. A mean value additionally specifies the input ensemble and whether it is conditioned on success. Detection efficiency, conditional fidelity, heralding and overall success probability are distinct experimental quantities.
Suppose a perfect Bell pair is used for quantum teleportation, but the sender cannot distinguish the and Bell-basis measurement outcomes and guesses their labels uniformly. Each of the four actual outcomes has probability . Either ambiguous outcome gives an incorrect Pauli Z gate correction with conditional probability , so the total phase-flip probability is . The unconditional output is therefore the displayed phase-flip channel. For , its teleportation fidelity is . It ranges from on the Bloch sphere equator to at its poles. This averages the output density operators, including the ambiguous rounds.
Temperature scaling 2026-11-05
Temperature scaling is a method of probability calibration that divides a model's neural network logits by one fitted positive scalar before applying the softmax function. The temperature is selected after model training, using labels in a held-out validation set to minimize the negative log-likelihood. It adjusts reported probabilities without retraining all model weights.[ref]
For class scores and ,
For the synthetic two-class scores , the first-class probability is
The selected class remains the first while its reported confidence decreases.
The preservation of the selected class follows directly: dividing by positive preserves every score ordering and equality, and exponentiation is strictly increasing. Thus the set of classes attaining the largest score is unchanged. This concerns ordering between classes within one example; it does not assert preservation of confidence ordering between different examples.
A motivating measurement in the original study compares reliability diagrams for a 110-layer image-classification network. Its reported expected calibration error changes from before calibration to after temperature scaling. This is a result for that experimental setup and binned metric, not a guarantee for every classifier or input distribution.[ref]
Use separate data for fitting the temperature and for final evaluation. One scalar can improve an overall confidence mismatch but cannot represent every class-specific or input-dependent error. The name refers to this probability transformation; no physical temperature is being measured.
Ternary prefix code Created 2026-10-25 Updated 2026-10-31
A prefix code whose codewords use a three-symbol alphabet, conventionally . No complete codeword is a prefix of another, permitting sequential decoding by following a code tree to a tree leaf. Its expected codeword length is trits per source symbol. For source probabilities , the codewords give expected codeword length trits.
Tetrahedral qubit POVM Created 2026-10-14 Updated 2026-11-05
Choose the four unit Bloch vectors at the vertices of a regular tetrahedron, so and . The displayed four positive rank-one effects sum to . On input , their probabilities are , giving the reconstruction . This realizes the smallest fixed qubit informationally complete POVM.
A bead in a liquid keeps undergoing Brownian motion even when an optical trap holds its average position fixed. For a stationary harmonic optical trap in thermal equilibrium at absolute temperature , the equipartition theorem assigns mean potential energy to this one quadratic coordinate. With the Boltzmann constant and at equilibrium,
At the same temperature, quadrupling the spring constant divides the position variance by four and the standard deviation by two. The Boltzmann factor gives the complete equilibrium probability density function, a Gaussian distribution:
The prefactor makes ; its second moment is . The probability density function has units of inverse length, and its area gives probability. This density is exact for the ideal quadratic model. In a finite optical trap it approximates the occupied well when thermal excursions remain within the harmonic region and escape is negligible during observation. Driven motion, appreciable nonconservative forces or laser heating need not give this equilibrium density.
Figure 1.
Thermal force calibration model: exact Gaussian position densities for relative trap stiffnesses 1, 2 and 4 at a common temperature
. Details.
Here , and . The plotted density per unit is . Each curve has total area one; their dimensionless standard deviations are , and . The curves are exact model densities, not noisy experimental data.
Brownian motion becomes a measuring tool: a calibrated position measurement supplies , and the known temperature supplies . Once is known, a slowly applied constant external force shifts the mean by , so in the same harmonic range. Fluctuations therefore calibrate the conversion from displacement to force. Svoboda and Block used this equipartition theorem method when calibrating their bead traps. [ref].
For the example stiffness and , and . Using a variance measured in uncalibrated detector units would not determine that physical stiffness. Measurement noise, finite exposure time, drift and insufficient sampling of the correlated Brownian motion can bias the estimate; record positions over many optical trap relaxation times and account for the detector response. The deterministic relaxation curve above describes relaxation without thermal forcing; an individual thermal trajectory continues to fluctuate.