notes

uni notes
git clone git://popovic.xyz/notes.git
Log | Files | Refs

todo.md (1305B)


      1 TODO:
      2 
      3 Structure:
      4 * introduction to problem
      5     * What are we trying to solve
      6 * Background
      7     * Coding
      8     * Neural Networks
      9     * Decomposition Cases
     10     * Gauss Newton and Landweber
     11     * Tikhonov Regularization
     12     * Space Regularization
     13     * symmetries
     14 * Theoretical Results
     15     * Gauss-Newton type method for problem
     16     * Convergence of Gauss-Newton
     17     * Newton's method with nn operator and linear independence
     18     * Results of linear independence
     19 * Experimental Results
     20     * Gauss-Newton
     21     * Landweber
     22     * Circular NNs with result.
     23 
     24 We need
     25     * more on coding
     26     * Tikhonov regularization
     27     * Space regularization
     28     * more on the decomposition cases
     29     * Proof of local convergence of GN
     30     * Proof of linear indepencdence thm
     31     * Proof of GN convergence with linear independence
     32     * Reproduction of numerical results of GN
     33     * Reproduction of numerical results of landweber
     34     * Reproduction of numerical results of circualr networks.
     35     * (reproduction or more in depth explanations)
     36 
     37 
     38 Proving convergence:
     39     * prove Lipschitz-Differentiable immersion of shallow NNs
     40     * Linear independence of the activation function, first derivative and first
     41         moment of the first derrivaive
     42     * Newton Minkowski conditions for shallow NNs
     43     * Moore Penrose inverse
     44 
     45 
     46