richard-notes.txt

Functions/methods from python libraries that are needed:

To implement scipy.optimize
curve_fit(f, xdata, ydata, p0=None, sigma=None, absolute_sigma=False, check_finite=True, bounds=(-np.inf, np.inf), method=None, jac=None, **kwargs)

* called with curve, xs and ys, returns params, covr but only params used
* other parameters at defaults
* a lot of checking/setup code can be removed - setup based of default value paths
* calls leastsq(func, x0, args=(), Dfun=None, full_output=0,
    col_deriv=0, ftol=1.49012e-8, xtol=1.49012e-8,
    gtol=0.0, maxfev=0, epsfcn=None, factor=100, diag=None)
    - only func, x0 and Dfun set (Dfun == None in default params path of curve_fit)
    - hits many matrix functions from numpy/scipy
        . _minpack._lmdif
        . _minpack._lmder
        . take, eye, transpose, tri/tril/triu, dot, inv, ...
        . numpy.sum, finfo/iinfo - machine limits for data types,
        . max of values in array
    - calls svd - from decomp_svd
    
* things like _check_func not needed, as we are in control of the shape of the matricies?

* What exceptions/errors to throw?


Debug of iris dataset run.

UMAP.fit_transform(X, y)
  - X : 2D ndarray of float64 (150,4)
  - y : None

find_ab_params(spread, min_dist)
  - spread : 1.0
  - min_dist : 0.1
curve_fit()
  - always called with default settings, so always returns same values