Basis functions are the building blocks of a function space in Linear Algebra and Differential Equations. Any function in that space can be written as a linear combination of them, which is why they show up in least squares approximations.
Basis functions are the specific functions you choose to represent other functions in a Linear Algebra and Differential Equations problem. If they are linearly independent and span the function space you care about, then any function in that space can be written as a linear combination of them.
That idea should feel familiar if you have seen basis vectors in linear algebra. It is the same structure, just moved from vectors in \u211d^n to functions. Instead of writing a vector as a combination of column vectors, you write a function as a combination of functions like 1, x, x^2, sin(x), or e^x, depending on the problem.
In least squares, basis functions give you a controlled way to approximate data or a complicated target function. You pick a set of basis functions, then solve for the coefficients that make the linear combination fit as closely as possible. The result is not usually an exact match, but it is the best fit within the span of the chosen basis.
The choice of basis matters. Polynomial basis functions are common for trends that curve smoothly, while trigonometric basis functions work well for periodic patterns, and exponentials show up in growth and decay models. A poor choice can make the approximation awkward or inaccurate, even if the math is done correctly.
Orthogonal basis functions make life easier because the functions do not overlap in the same way. When basis functions are orthogonal, projection and least squares calculations simplify, and the coefficients are easier to interpret. That is one reason orthogonal bases come up so often right next to the normal equations and projection ideas in this topic.
A good way to think about basis functions is as your model's ingredients. The coefficients tell you how much of each ingredient to use, and the span tells you what kinds of functions your recipe can produce.
Basis functions are the bridge between abstract vector space ideas and the actual approximation problems you solve in Linear Algebra and Differential Equations. They let you turn a hard function-fitting problem into a linear algebra problem with coefficients, matrices, and projections.
That shows up directly in least squares approximations. Instead of asking for a perfect solution, you ask for the best approximating vector or function inside the span of your chosen basis. Once you choose the basis, the rest becomes a coefficient problem, often organized with a design matrix and solved through normal equations or projection methods.
This term also explains why two different models can fit the same data differently. A polynomial basis may capture curvature, while a trigonometric basis may capture repeating behavior. If you understand basis functions, you can justify why one setup makes more sense than another instead of just trying formulas at random.
The term matters later in differential equations too, because many solution methods represent solutions as combinations of simpler functions. That habit of building solutions from basis functions is a core way the course connects structure, computation, and approximation.
Keep studying Linear Algebra and Differential Equations Unit 6
Visual cheatsheet
view galleryLinear Combination
Basis functions matter because you use them in a linear combination. The coefficients are the weights that tell you how much of each basis function appears in the final approximation. If the functions are linearly independent, that combination gives you a clean way to build and compare models.
Function Space
A basis only makes sense relative to a function space. You are not choosing random functions, you are choosing functions that span the space you care about, such as polynomials of a certain degree or functions used in a least squares model. The space tells you what counts as reachable.
Orthogonality
Orthogonality makes basis functions easier to work with because the functions do not interfere with one another the way non-orthogonal functions do. In least squares, orthogonal bases simplify projections and often make coefficient calculations cleaner. That is why orthogonal choices are so popular in approximation problems.
least squares method
The least squares method is where basis functions get used in a practical way. You choose a set of basis functions, then find the coefficients that minimize the total squared error between the model and the data. The method turns approximation into a structured linear algebra problem.
A problem set or quiz question may give you a set of candidate functions and ask whether they can serve as basis functions, or it may ask you to build the least squares model from them. Your job is to check linear independence, identify the span, and decide whether the chosen functions can represent the target behavior. If the basis is orthogonal, you may be asked to use that fact to simplify projections or coefficient calculations.
You will also see basis functions when a question asks you to explain why one approximation works better than another. A good answer connects the function choice to the pattern in the data, like using polynomials for smooth curvature or sines and cosines for repeating motion. In differential equations, the same idea shows up when a solution is written as a combination of simpler functions rather than one closed-form expression.
Basis functions and basis vectors are the same idea in different settings. Basis vectors span ordinary vector spaces like \u211d^n, while basis functions span function spaces. In this course, you move between them when least squares is written either as vectors in matrix form or as functions in an approximation problem.
Basis functions are the functions that generate a function space through linear combinations.
In least squares, you choose basis functions first, then solve for the coefficients that give the best fit.
A good basis is linearly independent and suited to the pattern in the data or equation.
Orthogonal basis functions are easier to work with because they simplify projections and coefficient finding.
The term connects linear algebra ideas like span and dimension to function approximation in differential equations.
Basis functions are the set of functions you use to build every function in a chosen function space. In this course, they show up when you approximate data or solve a problem by writing the answer as a linear combination of simpler functions. They are the function-space version of a basis in linear algebra.
They are the same idea, but in different spaces. Basis vectors span vector spaces like \u211d^n, while basis functions span spaces of functions. The vocabulary changes, but the logic is still about linear independence, span, and coordinates or coefficients.
You pick a set of basis functions, then find the coefficients that make their linear combination as close as possible to the target data. The best fit minimizes squared error, so the basis functions determine what kind of model you are allowed to build. The final approximation depends heavily on that choice.
Polynomials are a very common example, especially in smooth curve fitting. Depending on the problem, sine and cosine functions or exponentials may be better choices. The right basis depends on whether your data looks curved, periodic, or like growth and decay.