Data Science

3 Credits

Contact: Frédéric CHAMBAT

Any set of data is a signal. In the sciences of the universe, data are sometimes provided in large amounts and processing them is a challenge in itself. But even in simple cases, it is useful to know the fundamentals: the notion of frequency, decomposition on a base of sinusoids, a Fourier spectrum etc.

Temperature versus time (current = 0) recorded in the Dome C ice core. Signal processing allows, among other things, to identify periodicities in this type of curve.

Contents:

  • Fourier II (18h)
      • Dirac distribution (2h)
      • Fourier transform, convolution, TF of usual functions, diff. eq. (6h)
      • Signal processing (10h)
        • Convolution and linear physics
        • Green's functions
        •  Truncation
        •  Discretization
        • DFT and FFT
        • Apodisation, detrending
        • Programming of these concepts (4 lab sessions)
  • AI / Data science (6h)
      • PCA (2h)
      • Introduction to classification (k-means method) (2h)
      • Introduction to neural networks (2h) 
en_GBEnglish