Rooster
HC: woensdag, 15:15-17:00, Ruppert 134
WG: vrijdag,
11:00-12:45, Ruppert 135,
op 30/9 BG 115
Tentamen en tussentoets: open boek.
Bijdragen tot het eindcijfer:
ANNOUNCEMENTS
Lecture 1: Artificial neurons and networks. Activation functions,
weights. Threshold linear activation function, a neuron computing binary "and"
function.
Vectors, matrices, operations with them. Vector space
Rn. Geometric interpretation of vectors. Elementary transformations and
Gaussian elimination method.
Lecture 2: Inner product of vectors and its properties. Geometric interpretation of solutions of a system of linear equations; lines in R2 and hyperplanes in Rn; normal vector to a hyperplane; linear subspaces of Rn; linear independence; testing linear independence; linear span of a set of vectors; dimension of a vector space.
The following problem was considered at the workgroup.
Problem. Given two different points A and
B in R2 find
an equation defining the straight line going through A and B.
Solution 1. The equation has the form ax+by=c. Substituting the
coordinates of A for x and y validates the equation, similarly for B. So, we obtain
a system of two linear equations with the
unknowns a, b and c. Solve this system and take any non-zero solution for the
sought a, b and c.
Solution 2. Vector (a,b) is orthogonal to the line defined by ax+by=c.
(Why?) Also: (-b,a) is orthogonal to (a,b). Now let (u,v) be the
coordinates of the vector B-A. Then we can take a=-v and b=u.
How can we now determine c?
Problem for the WG: solve the same question in an n-dimentional space, that is, find the system of equations for the straight line going through two given points A and B. (It is easier to generalize Solution 2 above.)
More WG exercises: .pdf (Hado).
Lecture 3: Rank of a matrix; equality of column and row ranks; geometric interpretation of rank. Basis of a space; the standard basis in Rn; coordinates of a vector in a given basis. The change of coordinates in Rn; transfer matrix from one basis to another. Linear transformations, their matrices and geometric interpretation. Inverse matrix; regular matrices are invertible and vice versa; proof of the inversion algorithm by elementary transformations.
Some WG exercises: .pdf (Lev).Lecture 4: Quadratic forms, their geometric interpretation. Matrix of a form. Transformation of matrices when coordinates are changed. Bringing a matrix to a canonical basis. Rank and index of a quadratic form.
Lecture 5: Derivatives of functions in one and several variables. Taylor series for functions in one variable. Partial derivatives. Gradient. Jacobi matrix.
Homework week 5: .pdf (Hado).Lecture 6: Total derivative. Derivatives of a composition of functions; chain formulas (K p. 444 + extra). Directional derivative. Gradient descent method. Second and higher order derivatives (K A59); Taylor formula up to second order. Local maxima and minima; investigation of extremum points (K p.990-993).
Homework week 6 .pdf (Hado).Lecture 7: Regression function. Linear regression. Least squares problem. Pseudoinverse matrix.
Some WG exercises: .pdf (Hado).Part 1. Linear Algebra