Skip to main content
Advanced mathematical modelling
Lesson 62 of 75
Easy practice

Inverse problems

Objective. Infer hidden parameters from indirect observations.

  1. 1. Learn
  2. 2. Worked example
  3. 3. Practice
  4. 4. Feedback
  5. 5. Continue

Learn: the key idea

Inverse problems infer hidden model parameters from indirect, noisy observations, which is often ill-posed because small data errors can cause large, unstable changes in the inferred parameter. Tikhonov regularisation stabilises this by minimising ‖Ax−b‖² + λ‖x‖² instead of just the data misfit, trading a little bias for much reduced sensitivity to noise, where λ controls the trade-off. For example, estimating a source strength from a noisy measurement by simple division can give wildly different answers for tiny noise changes, whereas adding a regularisation term damps this instability. Choosing λ, often via cross-validation, balances fitting the data against keeping the solution stable.

Worked example

An inverse problem estimates a parameter x from the model 2x = b with regularisation minimising (2x−b)² + λx². For b = 10 and λ = 1, find the regularised estimate of x using the normal equation (4+λ)x = 2b.

2210142
Bar chart of the values 2, 2, 10, 1, 4, 2
  1. 1Write the regularised normal equation (4+λ)x = 2b for this linear model.
  2. 2Substitute λ = 1 and b = 10: (4+1)x = 2(10), giving 5x = 20.
  3. 3Divide both sides by 5: x = 20/5.
  4. 4Simplify 20/5 = 4, the regularised parameter estimate.

Answer: 4

Concept mastery

Based on your past attempts at this lesson — the question types to practise again come first.

Complete a practice set to see your concept breakdown here.

Practice

Practice happens on its own screen, one question at a time. Answers stay hidden until you submit yours, and you can stop and pick up at the same question later.

Start practising

Common mistake. A common mistake is solving an ill-posed inverse problem by direct inversion of the forward operator without any regularisation, which amplifies measurement noise into wildly unstable parameter estimates.

Your progress

Tick each step as you finish it. Your exact reading position is saved automatically, so Continue drops you back on the same line.

0 of 5 steps complete

Your place saves automatically as you read.

My dashboard