Hessian matrix
In mathematics, the Hessian matrix, Hessian or (less commonly) Hesse matrix is a square matrix of second-order partial derivatives of a scalar-valued function, or scalar field.
≈ 8 min read
Hessian matrix. encyclopegia.org, snapshot enwiki-001375009356. /a/hessian-matrix
Names in other languages
No names in other languages are held for this entry yet.
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
In mathematics, the Hessian matrix, Hessian or (less commonly) Hesse matrix is a square matrix of second-order partial derivatives of a scalar-valued function, or scalar field. It describes the local curvature of a function of many variables. The Hessian matrix was developed in the 19th century by the German mathematician Ludwig Otto Hesse and later named after him.
Hesse originally used the term "functional determinants". The Hessian is sometimes denoted by H or \nabla\nabla or \nabla^2 or \nabla\otimes\nabla or D^2 . \mathbf H_f= \begin{bmatrix} \dfrac{\partial^2 f}{\partial x_1^2} & \dfrac{\partial^2 f}{\partial x_1\,\partial x_2} & \cdots & \dfrac{\partial^2 f}{\partial x_1\,\partial x_n} \\[2.2ex] \dfrac{\partial^2 f}{\partial x_2\,\partial x_1} & \dfrac{\partial^2 f}{\partial x_2^2} & \cdots & \dfrac{\partial^2 f}{\partial x_2\,\partial x_n} \\[2.2ex] \vdots & \vdots & \ddots & \vdots \\[2.2ex] \dfrac{\partial^2 f}{\partial x_n\,\partial x_1} & \dfrac{\partial^2 f}{\partial x_n\,\partial x_2} & \cdots & \dfrac{\partial^2 f}{\partial x_n^2} \end{bmatrix}.
That is, the entry of the th row and the th column is (\mathbf H_f)_{i,j} = \frac{\partial^2 f}{\partial x_i \, \partial x_j}. If furthermore the second partial derivatives are all continuous, the Hessian matrix is a symmetric matrix by the symmetry of second derivatives. The determinant of the Hessian matrix is called the .
The Hessian matrix of a function f is the transpose of the Jacobian matrix of the gradient of the function f ; that is: \mathbf{H}(f(\mathbf{x})) = \mathbf{J}(\nabla f(\mathbf{x}))^T. If f is a homogeneous polynomial in three variables, the equation f = 0 is the implicit equation of a plane projective curve. The inflection points of the curve are exactly the non-singular points where the Hessian determinant is zero.
It follows by Bézout's theorem that a cubic plane curve has at most 9 inflection points, since the Hessian determinant is a polynomial of degree 3. y = f(\mathbf{x} + \Delta\mathbf{x})\approx f(\mathbf{x}) + \nabla f(\mathbf{x})^\mathsf{T} \Delta\mathbf{x} + \frac{1}{2} \, \Delta\mathbf{x}^\mathsf{T} \mathbf{H}(\mathbf{x}) \, \Delta\mathbf{x} where \nabla f is the gradient \left(\frac{\partial f}{\partial x_1}, \ldots, \frac{\partial f}{\partial x_n}\right). Computing and storing the full Hessian matrix takes \Theta\left(n^2\right) memory, which is infeasible for high-dimensional functions such as the loss functions of neural nets, conditional random fields, and other statistical models with large numbers of parameters.
For such situations, truncated-Newton and quasi-Newton algorithms have been developed. The latter family of algorithms use approximations to the Hessian; one of the most popular quasi-Newton algorithms is BFGS.
Such approximations may use the fact that an optimization algorithm uses the Hessian only as a linear operator \mathbf{H}(\mathbf{v}), and proceed by first noticing that the Hessian also appears in the local expansion of the gradient: \nabla f (\mathbf{x} + \Delta\mathbf{x}) = \nabla f (\mathbf{x}) + \mathbf{H}(\mathbf{x}) \, \Delta\mathbf{x} + \mathcal{O}(\|\Delta\mathbf{x}\|^2) Letting \Delta \mathbf{x} = r \mathbf{v} for some scalar r, this gives \mathbf{H}(\mathbf{x}) \, \Delta\mathbf{x} = \mathbf{H}(\mathbf{x})r\mathbf{v} = r\mathbf{H}(\mathbf{x})\mathbf{v} = \nabla f (\mathbf{x} + r\mathbf{v}) - \nabla f (\mathbf{x}) + \mathcal{O}(r^2), that is, \mathbf{H}(\mathbf{x})\mathbf{v} = \frac{1}{r} \left[\nabla f(\mathbf{x} + r \mathbf{v}) - \nabla f(\mathbf{x})\right] + \mathcal{O}(r) so if the gradient is already computed, the approximate Hessian can be computed by a linear (in the size of the gradient) number of scalar operations. (While simple to program, this approximation scheme is not numerically stable since r has to be made small to prevent error due to the \mathcal{O}(r) term, but decreasing it loses precision in the first term. ) Notably regarding Randomized Search Heuristics, the evolution strategy's covariance matrix adapts to the inverse of the Hessian matrix, up to a scalar factor and small random fluctuations.
This result has been formally proven for a single-parent strategy and a static model, as the population size increases, relying on the quadratic approximation. A is used for the second-derivative test in certain constrained optimization problems.
Given the function f considered previously, but adding a constraint function g such that g(\mathbf{x}) = c, the bordered Hessian is the Hessian of the Lagrange function \Lambda(\mathbf{x}, \lambda) = f(\mathbf{x}) + \lambda[g(\mathbf{x}) - c] : \mathbf H(\Lambda) = \begin{bmatrix} \dfrac{\partial^2 \Lambda}{\partial \lambda^2} & \dfrac{\partial^2 \Lambda}{\partial \lambda \partial \mathbf x} \\ \left(\dfrac{\partial^2 \Lambda}{\partial \lambda \partial \mathbf x}\right)^{\mathsf{T}} & \dfrac{\partial^2 \Lambda}{\partial \mathbf x^2} \end{bmatrix} = \begin{bmatrix} 0 & \dfrac{\partial g}{\partial x_1} & \dfrac{\partial g}{\partial x_2} & \cdots & \dfrac{\partial g}{\partial x_n} \\[2.2ex] \dfrac{\partial g}{\partial x_1} & \dfrac{\partial^2 \Lambda}{\partial x_1^2} & \dfrac{\partial^2 \Lambda}{\partial x_1\,\partial x_2} & \cdots & \dfrac{\partial^2 \Lambda}{\partial x_1\,\partial x_n} \\[2.2ex] \dfrac{\partial g}{\partial x_2} & \dfrac{\partial^2 \Lambda}{\partial x_2\,\partial x_1} & \dfrac{\partial^2 \Lambda}{\partial x_2^2} & \cdots & \dfrac{\partial^2 \Lambda}{\partial x_2\,\partial x_n} \\[2.2ex] \vdots & \vdots & \vdots & \ddots & \vdots \\[2.2ex] \dfrac{\partial g}{\partial x_n} & \dfrac{\partial^2 \Lambda}{\partial x_n\,\partial x_1} & \dfrac{\partial^2 \Lambda}{\partial x_n\,\partial x_2} & \cdots & \dfrac{\partial^2 \Lambda}{\partial x_n^2} \end{bmatrix} The second derivative test consists here of sign restrictions of the determinants of a certain set of n - m submatrices of the bordered Hessian. Intuitively, the m constraints can be thought of as reducing the problem to one with n - m free variables.
(For example, the maximization of f\left(x_1, x_2, x_3\right) subject to the constraint x_1 + x_2 + x_3 = 1 can be reduced to the maximization of f\left(x_1, x_2, 1 - x_1 - x_2\right) without constraint.) Specifically, sign conditions are imposed on the sequence of leading principal minors (determinants of upper-left-justified sub-matrices) of the bordered Hessian, for which the first 2 m leading principal minors are neglected, the smallest minor consisting of the truncated first 2 m + 1 rows and columns, the next consisting of the truncated first 2 m + 2 rows and columns, and so on, with the last being the entire bordered Hessian; if 2 m + 1 is larger than n + m, then the smallest leading principal minor is the Hessian itself.
There are thus n - m minors to consider, each evaluated at the specific point being considered as a candidate maximum or minimum. A sufficient condition for a local is that these minors alternate in sign with the smallest one having the sign of (-1)^{m+1}.
A sufficient condition for a local is that all of these minors have the sign of (-1)^m. (In the unconstrained case of m=0 these conditions coincide with the conditions for the unbordered Hessian to be negative definite or positive definite respectively). \mathbf f(\mathbf x) = \left(f_1(\mathbf x), f_2(\mathbf x), \ldots, f_m(\mathbf x)\right), then the collection of second partial derivatives is not a n \times n matrix, but rather a third-order tensor. This can be thought of as an array of m Hessian matrices, one for each component of \mathbf{f} : \mathbf H(\mathbf f) = \left(\mathbf H(f_1), \mathbf H(f_2), \ldots, \mathbf H(f_m)\right).
This tensor degenerates to the usual Hessian matrix when m = 1. \operatorname{Hess}(f) \in \Gamma\left(T^*M \otimes T^*M\right) \quad \text{ by } \quad \operatorname{Hess}(f) := \nabla \nabla f = \nabla df, where this takes advantage of the fact that the first covariant derivative of a function is the same as its ordinary differential. Choosing local coordinates \left\{x^i\right\} gives a local expression for the Hessian as \operatorname{Hess}(f)=\nabla_i\, \partial_j f \ dx^i \!\otimes\! dx^j = \left(\frac{\partial^2 f}{\partial x^i \partial x^j} - \Gamma_{ij}^k \frac{\partial f}{\partial x^k}\right) dx^i \otimes dx^j where \Gamma^k_{ij} are the Christoffel symbols of the connection.
Other equivalent forms for the Hessian are given by \operatorname{Hess}(f)(X, Y) = \langle \nabla_X \operatorname{grad} f,Y \rangle \quad \text{ and } \quad \operatorname{Hess}(f)(X,Y) = X(Yf)-df(\nabla_XY).
Text derived from https://en.wikipedia.org/w/index.php?oldid=1375009356 under CC BY-SA 4.0; modifications have been made.