Back to papers
    bayes-theoremprobabilityhistory-of-mathematicsstatistics

    An Essay towards Solving a Problem in the Doctrine of Chances

    Thomas Bayes
    📅 1763🏛 Philosophical Transactions of the Royal Society of London (https://doi.org/10.1098/rstl.1763.0053)
    Problem

    Prior to this publication, there was no mathematical method to perform inverse probability, meaning estimating the probability of an unknown cause given observed events or data trials.

    Method

    Bayes proposed a physical and mathematical model of balls thrown on a table, proving that the probability of an unknown parameter lies within an interval by updating its uniform prior distribution based on the outcomes of success/failure trials.

    Finding

    The derivation of the basic formulation of posterior distributions, laying the foundational framework for what is now known as Bayesian inference.

    Limitations

    The paper assumes a uniform prior distribution for the unknown probability parameters, which sparked centuries of philosophical debate regarding subjective priors.

    Practical application

    Forms the mathematical backbone of Naive Bayes classifiers, Bayesian neural networks, state estimation algorithms (e.g., Kalman and Particle filters) in self-driving vehicles and robotic arms.

    📇 Summary flashcard — 13 analytical fields for this paper

    خلاصه

    The founding document of Bayesian statistics, outlining how to determine the probability of a cause based on observed effects.

    نمای سریع

    The origins of Bayesian updating and inverse probability.

    یافته‌های کلیدی

    The derivation of the basic formulation of posterior distributions, laying the foundational framework for what is now known as Bayesian inference.

    هدف

    To mathematically formulate how to update beliefs or parameters when new experimental evidence is gathered.

    روش

    Bayes proposed a physical and mathematical model of balls thrown on a table, proving that the probability of an unknown parameter lies within an interval by updating its uniform prior distribution based on the outcomes of success/failure trials.

    نتایج

    Introduced the mathematical foundation showing posterior probability is proportional to the likelihood multiplied by the prior.

    نتیجه‌گیری

    Unobserved probability parameters should be treated as random variables subject to revision based on empirical trials.

    مفاهیم کلیدی

    bayes-theorem، probability، history-of-mathematics، statistics

    مطالعه‌ی بیشتر

    https://doi.org/10.1098/rstl.1763.0053

    تحلیل

    This posthumously published paper remains one of the most influential scientific texts of all time, framing modern machine learning as a process of Bayesian updates.

    محدودیت‌ها

    The paper assumes a uniform prior distribution for the unknown probability parameters, which sparked centuries of philosophical debate regarding subjective priors.

    کارهای آینده

    Extension to continuous variables and complex prior formulations, which were subsequently developed by Pierre-Simon Laplace.

    کاربرد عملی

    Forms the mathematical backbone of Naive Bayes classifiers, Bayesian neural networks, state estimation algorithms (e.g., Kalman and Particle filters) in self-driving vehicles and robotic arms.

    References (this paper draws on) (0)

    No references recorded.

    Cited by (0) ▶

    No paper cites this one yet.

    Prerequisite learning path for this paper