testing testing
mathematicalmichael opened this issue · comments
#9 gets us started with a baseline.
-
add badge to README for codecov
-
get coverage above 50%
-
make final release with commented out code, then delete it pre-
v0.1
-
ignore plotting for now,
just make sure it runsskipping for coverage
-
test
funs
-
test that numerical solutions agree for some test problems
-
test
plot
-
test
util
-
test
norm
-
test against numpy basics with identity cov.
-
test that diagonals getting larger actually shrinks the evaluation
-
test problem that you have analytical solutions for
-
will be used in numerical comparisons
map_sol matches from sklearn.linear_model import Ridge
with w
and alpha
playing the same roles.
r = Ridge(alpha=1, fit_intercept=False).fit(X, y)
map_sol = mf.map_sol(X, np.zeros(100), y, w=1).ravel()
print(r.coef_ - map_sol)
next up: how does sample_weights
correspond to covariance / prior evaluation.