Solving Rational Equations

A rational equation contains rational expressions on at least one side. Solve by clearing denominators, then check candidate solutions against the original equation’s domain.

The procedure

1. Identify domain restrictions by listing values that zero any denominator.
2. Find the LCD (least common denominator) of all rational expressions.
3. Multiply both sides of the equation by the LCD. This clears all fractions.
4. Solve the resulting polynomial equation.
5. Verify each candidate against the domain. Discard any that match an excluded value (extraneous roots).

Example

Domain: $x\ne 0$. LCD $=4x$. Multiply both sides:

Verify: $x=4$ is in the domain. Check: $\frac{1}{4}+\frac{1}{2}=\frac{1}{4}+\frac{2}{4}=\frac{3}{4}$.

Cross-multiplication

For the simple form $\frac{a}{b}=\frac{c}{d}$, you can cross-multiply:

This is just multiplying both sides by $bd$ and cancelling.

Extraneous solutions — the trap

Multiplying both sides of an equation by a variable expression can introduce candidates that solve the cleared equation but not the original. Always verify.

Example: $\frac{2}{x-1}=\frac{x+1}{x-1}$. Multiply both sides by $(x-1)$:

But $x=1$ makes both denominators zero in the original equation. The “solution” is extraneous; the original equation has no solution.

Why this matters for ML

Many ML formulas have division (softmax denominators, Bayes marginals, normalization constants). The corresponding numerical-stability discipline:

• Don’t divide by a near-zero number. Add an epsilon (1e−8) to any denominator that could approach zero.
• Don’t take log(0). Same epsilon trick: log(p + 1e−8).
• Verify your formula against edge cases. What happens at x = 0? At very large x?

The “always check the domain” habit you build with rational equations transfers directly to “always check for NaN propagation” in ML training loops.