Skip to the content ForoHackers
Log inJoin

Linear algebra: start here

Linear algebra
#631 · Ophuscado · · Report

Linear algebra studies vectors, matrices and the linear maps between them. It is the main tool for anything with many numbers at once: graphics transform points with matrices, search ranks pages with eigenvectors, and machine learning is largely matrix multiplication. Learn it as geometry first and calculation second.

Key concepts

  • Vectors and matrices: a vector is a point or an arrow; a matrix is a linear map, and multiplying matrices composes maps.
  • Linear systems: solving Ax = b by elimination, and knowing when there is no solution, one, or infinitely many.
  • Vector spaces: span, independence, basis and dimension, the structure behind every solution set.
  • Eigenvalues and eigenvectors: directions a map only stretches; they reveal its long-run behaviour.
  • Decompositions: LU, QR and the singular value decomposition factor a matrix into easier pieces.

Vocabulary

  • Rank: the number of independent columns, the true dimension of what a matrix can reach.
  • Orthogonal: perpendicular; orthogonal matrices rotate and reflect without distorting lengths.
  • Determinant: a number that says how a map scales volume, and whether it can be inverted.
  • Least squares: the best approximate solution of a system with no exact one.

What to study, in order

  1. Vectors, dot product and the geometry of lines and planes; draw them.
  2. Implement Gaussian elimination and use it to solve systems and invert small matrices.
  3. Study span, basis, rank and null space with small examples.
  4. Compute eigenvalues for 2 by 2 matrices by hand, then use a numerical library for larger ones.
  5. Fit a line by least squares and compress an image with the SVD.

Related tags

#numerical-analysis #machine-learning #computer-graphics #discrete-math #signal-processing

Log in or join to reply.