Volunteer teaching / ENSEEIHT

Volunteer teaching

I teach as a volunteer at ENSEEIHT, alongside Hamza Kaouach and Jean-Yves Tourneret. In materials and semiconductor physics, I tutor students on band structures and the dielectric and magnetic response of materials, and I wrote a reference document with worked solutions for tutorial instructors. I also gave a lecture to my own class on likelihood and statistical estimation, built around a physical problem: estimating the height of ocean waves from a radar altimeter echo.

FIG. 01 · Radar altimeter over a rough seaIllustration.
01

Materials and semiconductor physics

  • SettingTutorials around the materials physics course taught by Hamza Kaouach in the electrical engineering programme at ENSEEIHT.
  • TopicsSemiconductor band structures, dielectric and magnetic response of materials and crystallography, with applications to components, insulation and sensors.
  • Reference document172 pages for tutorial instructors: electromagnetism reminders, dielectric and magnetic materials, and fully worked solutions of the first two tutorial sheets, with 73 figures and 62 numerical checks.
02

Lecture: likelihood and statistical estimation

I gave a lecture to my own class on likelihood and statistical estimation. The worked example came from satellite oceanography: how a noisy altimeter echo, compared with a physical model, yields the significant wave height of the sea. The example draws on published work by TéSA and CLS, presented as a case study.

Example: wave height from a radar echo

Estimating significant wave height from a radar altimeter echoAn altimeter sends a short pulse to the sea surface. Over a calm sea the returns arrive together and the echo rises steeply; over a rough sea, crests return before troughs and the leading edge spreads in time. A noisy measured echo is compared with an echo model whose parameters include the epoch, the significant wave height and the amplitude; maximum likelihood or maximum a posteriori estimation gives the sea-state parameters. Curves are illustrations.Radar altimetershort pulse, nadirEcho model p(y | θ)θ: epoch, wave height,amplitudeEstimatormaximum likelihood or MAPEstimated sea statesignificant wave height,range, backscattercalm searough seacrests return before troughs:the echo spreads in timeECHO POWERTIME (RANGE GATES)epochcalm: steepleading edgerough: gentler slopedots: noisy echoILLUSTRATION · SIMPLIFIED MODEL, NOT MEASURED DATA
  • Data

FIG. 02From echo to sea state. Simplified illustration: the curves come from a textbook-style echo model and synthetic noise, not from satellite data.
  • MeasurementThe altimeter sends a short radar pulse straight down and records the returned power over time: the echo, or waveform. Every measured echo is noisy.
  • ModelOver the ocean, the echo follows the Brown model. Its parameters include the epoch, which gives the range, the amplitude, linked to the backscatter coefficient, and the significant wave height.
  • Sea stateOver a calm sea the returns arrive together and the leading edge rises steeply. Over a rough sea, wave crests return before troughs, so the leading edge spreads in time and its slope is gentler. This describes the sea surface, not the sea floor.
  • EstimationMaximum likelihood keeps the parameters that make the observed echo most probable. Adding prior information on the parameters gives a Bayesian estimate, such as the maximum a posteriori.
The estimators

\(y\) is the measured echo, sampled over time; \(\theta\) gathers the model parameters (epoch, significant wave height, amplitude).

\[\hat\theta_{\mathrm{ML}} = \arg\max_{\theta}\; p(y \mid \theta)\]\[p(\theta \mid y) \propto p(y \mid \theta)\, p(\theta)\]\[\hat\theta_{\mathrm{MAP}} = \arg\max_{\theta}\; \big[\log p(y \mid \theta) + \log p(\theta)\big]\]

\(p(y \mid \theta)\) is the likelihood, set by the echo model and its noise; \(p(\theta)\) is the prior, an assumption on the parameters such as their plausible range. The data pull the estimate towards the best fit; the prior steadies it when an echo is too noisy to decide alone.