Main content

### Course: Multivariable calculus > Unit 1

Lesson 6: Visualizing multivariable functions (articles)# Reduce reliance on graphs

Although graphs are a great way to think about single variable functions, they don't always work for multivariable functions.

## Graphs are not the only way

If you have a single-variable function, like the one below, it's common to visualize it using a graph.

However, it's important to remember that

*graphs are not the same thing as functions*. This might seem obvious, but graphs are so useful for representing single-variable functions that people often hold on to the idea of graphs a little too tightly as they shift focus to multivariable functions.For example, do you remember the derivative? At some point you may have seen the formal definition, which looks like this:

Be honest, though, how often did you actually think about this limit while doing exercises, learning how to compute derivatives, and interpreting the meaning of derivatives?

It's much simpler to think of the derivative as representing the slope of the graph of $f$ . And there's nothing wrong with that! At least, for single-variable calculus there's not.

In multivariable calculus, we will not always visualize functions with graphs. As a result, when we extend the idea of a derivative, you cannot always think about it as a slope. But that doesn't mean we won't visualize it! It's just that the visualization might be somewhat different at times.

And, in the same way, the understanding of an integral as computing the signed area under a curve is so useful that students in single-variable calculus rarely think about it differently. Why would you? Don't fix what isn't broken, right?

Mastering multivariable calculus requires the flexibility to think of functions differently—still visually, but differently. It also requires incorporating fundamental notions like derivatives and integration into those new ways of thinking.

For example, the derivative is fundamentally asking how the output of a function changes as you slightly tweak its input. If a function has a multidimensional output, interpreting this as "slope" doesn't really make sense. Instead, you might have to visualize how a small change to the input influences each coordinate of the output.

Similarly, integration is fundamentally adding up a bunch of tiny little values—well, infinitely many infinitely tiny values—but this does not always mean area. In physics, for example, it's common to compute the "work" done on an object by some force using an integral, but there's not always a clear way to view that "work" as any kind of area.

## Five different visualizations

In the next several articles, we'll go through five different ways to visualize multivariable functions. Here we'll just get a taste for each one.

In each of the following descriptions, "$(x,y)$ , like $(2,5)$ , and outputs a single number, like $5$ , the input space would be the $xy$ -plane and the output space would be the real number line.

**input space**" and "**output space**" refer to where the input and output of a function live. For example, if a function takes in an ordered pair**Graphs, our old friend.**Graphs have the benefit of showing both the input space and the output space at once, but as a result, they are highly limited by dimension. For this reason, they are only really useful for single-variable functions and multivariable functions with a two-dimensional input and a one-dimensional output.**Contour maps.**Contour maps only show the input space and are useful for functions with a two-dimensional input and a one-dimensional output.**Parametric curves/surfaces.**Parametric curves and surfaces only show the output space and are used for functions whose output space has more dimensions than the input space.**Vector fields.**These apply to functions whose input space and output space have the same number of dimensions. For example, functions with two-dimensional inputs and two-dimensional outputs, or three-dimensional inputs and three-dimensional outputs can be used with vector fields.**Transformations.**These have the benefit of applying to any function, no matter the dimension of the input and output space. However, the downside is that they can only be represented using an animation or a schematic drawing. As such, they are most useful for gaining a conceptual understanding of what a function is doing, but are impractical for representing the function precisely.

With each new topic and definition that you learn, a good way to test your understanding is to see if you can make sense of it in the context of functions that you visualize in each of these different ways. For example, the derivative indicates slope in the context of graphs, but the multivariable version of a derivative might mean something entirely different for parametric functions, vector fields, and contour maps.

## Want to join the conversation?

- Can someone name some examples of the different dimensions of output and input? I haven't grasped the differences.(4 votes)
- Temperature by location would be an input with 2 dimensions (a point), and a 1 dimensional output, a single scalar value.(3 votes)

- In this article it says "But that doesn't mean we won't will visualize it!".. Do I get kudos for finding this mistake and for proving that I am actually reading these articles? :)(14 votes)
- Is the following statement (taken from the above explanation) correct?

"Contour maps only show the input space and are useful for functions with a two-dimensional input and a one-dimensional output."

I believe that it should read as follows:

"Contour maps show selected points in the output spaces and corresponding input spaces and are useful for functions with a two-dimensional input and a one-dimensional output."(3 votes) - what will be the best visualization tool for above Five different visualizations?(3 votes)
- this statement :

"vector field can only be used on same dimensional input and output"

the way i see it, input dimension can be as many as we want while the output is 2 or 3 dimensional. do i get the concept wrong or is that statement is a rule decided by people?(2 votes)- I don't think the input can be n-dimensional as the vector(output) stems from the point(input). This implies that both of them are in the same space.

Thus, the statement that you make :*"vector field can only be used on same dimensional input and output"*

is the correct one.

If it's still not clear, I recommend you watch this:

https://www.khanacademy.org/math/multivariable-calculus/thinking-about-multivariable-function/visualizing-vector-valued-functions/v/vector-fields-introduction(3 votes)

- what are other examples of ways to visualize functions?(3 votes)
- So in f(x,y)= [some fxn with x and y] = z, I always thought the output was 3D, but it the output actually only one dimension since the output is Z?(1 vote)
- What software I can use to make a transformation animation?(1 vote)
- these dont have corresponding videos, right?(1 vote)
- were is the questions and the activity(1 vote)
- Sadly there are no questions and activities.

If you would like some problems with solutions, I would be happy to post a link where you can get them.(1 vote)