How to learn a math formula: derive it, compress it, then check it

A formula is a short statement that hides definitions, algebra, and conditions. The rule $\frac{d}{dx}x^2=2x$ fits on one line, but it depends on what a derivative means and on a limiting process. If the line is learned without that structure, a small variation can feel like an entirely new rule.

The first pass is therefore not to memorise every symbol. It is to ask four questions: what does the formula calculate, which quantities enter it, which conditions make it valid, and what feature of the derivation should make the result plausible? These questions turn the formula from an instruction into an argument that can be reconstructed.

For $f(x)=x^2$, the derivative is defined by the difference quotient $f'(x)=\lim_{h\to0}\frac{(x+h)^2-x^2}{h}.$ Expanding the square gives $(x+h)^2=x^2+2xh+h^2$, so the numerator becomes $2xh+h^2$. For $h\ne0$, we can divide by $h$: $\frac{2xh+h^2}{h}=2x+h.$ Taking the limit as $h$ approaches zero leaves $f'(x)=2x$.

One detail is worth keeping visible. We cancel $h$ while $h$ is non-zero; only after that simplification do we take the limit toward zero. The derivation explains both the coefficient $2$ and the remaining factor $x$. Neither appears by a mysterious movement of symbols.

Not every line deserves equal space in memory. In this derivation, the decisive step is the expansion $(x+h)^2-x^2=2xh+h^2.$ It shows how a small input change $h$ produces a first-order change $2xh$ plus a smaller squared term $h^2$. After division by $h$, the expression is $2x+h$; the part that remains as $h$ tends to zero is $2x$.

This is the structural cue to keep: expand the nearby value, subtract the original, divide by the change, then take the limit. If a sign or coefficient is forgotten, that sequence can recover the rule. Remembering the purpose of the difference quotient is more useful than trying to preserve a photograph of every line.

A useful memory note can be much shorter than the full derivation. Write the purpose, the cue, the result, and one condition: purpose: instantaneous rate of change; cue: expand the difference quotient; result: $2x$; condition: cancel $h$ only for $h\ne0$, then take the limit. That is enough information to rebuild the argument.

This kind of compression is different from copying only the final formula. It preserves the point at which an error is most likely. The same principle applies elsewhere: a quadratic formula note should keep the standard form and the role of the discriminant; a percentage-change note should keep the original value in the denominator and state that it must be non-zero.

After reading the derivation, hide it and reproduce it for $x^2$. Then change one feature. For $g(x)=(x-1)^2$, the same difference-quotient process gives $g'(x)=2(x-1).$ For $p(x)=5x^2$, the constant factor remains attached, giving $p'(x)=10x$. These are close enough to use the same structure, but different enough to reveal whether the cue is understood.

The aim is not to generate many nearly identical exercises. Two or three deliberate variations are enough to test the boundaries of the remembered rule. If the reconstruction stalls, return to the first line you trust and identify the missing decision rather than immediately rereading the whole solution.

At $x=3$, the formula predicts a slope of $2(3)=6$. A nearby secant with $h=0.01$ has slope $\frac{(3.01)^2-3^2}{0.01}=6.01.$ The secant slope is close to $6$ and moves toward it as $h$ becomes smaller. This numerical check does not replace the derivation, but it can expose a missing factor or an implausible sign.

A second check comes from the graph. The function $x^2$ decreases for negative $x$, is flat at $x=0$, and increases for positive $x$. The derivative $2x$ has exactly those signs. Good checks use information that was not simply copied from the same algebraic line.

You do not need to rederive a familiar formula every time you use it. Direct recall is efficient once you can state what the rule does, recognise its conditions, and produce a check. Re-derive when the formula has been partly forgotten, when a new variation changes its assumptions, or when two remembered versions appear to conflict.

A practical study session can therefore be short: derive once with the notes open, compress the reasoning into four cues, reproduce it without looking, solve one nearby variation, and finish with an independent check. You can ask Euler's tutor for a hint at any one of those stages instead of requesting the complete solution immediately. The related guide in Euler Learn gives the focused derivation and FAQs.