[Numpy-discussion] Matrix rank default tolerance - is it too low?
Matthew Brett
matthew.brett at gmail.com
Sun Jun 17 03:49:02 EDT 2012
Hi,
On Sat, Jun 16, 2012 at 1:33 PM, Matthew Brett <matthew.brett at gmail.com> wrote:
> Hi,
>
> On Sat, Jun 16, 2012 at 8:03 PM, Matthew Brett <matthew.brett at gmail.com> wrote:
>> Hi,
>>
>> On Sat, Jun 16, 2012 at 10:40 AM, Nathaniel Smith <njs at pobox.com> wrote:
>>> On Fri, Jun 15, 2012 at 4:10 AM, Charles R Harris
>>> <charlesr.harris at gmail.com> wrote:
>>>>
>>>>
>>>> On Thu, Jun 14, 2012 at 8:06 PM, Matthew Brett <matthew.brett at gmail.com>
>>>> wrote:
>>>>>
>>>>> Hi,
>>>>>
>>>>> I noticed that numpy.linalg.matrix_rank sometimes gives full rank for
>>>>> matrices that are numerically rank deficient:
>>>>>
>>>>> If I repeatedly make random matrices, then set the first column to be
>>>>> equal to the sum of the second and third columns:
>>>>>
>>>>> def make_deficient():
>>>>> X = np.random.normal(size=(40, 10))
>>>>> deficient_X = X.copy()
>>>>> deficient_X[:, 0] = deficient_X[:, 1] + deficient_X[:, 2]
>>>>> return deficient_X
>>>>>
>>>>> then the current numpy.linalg.matrix_rank algorithm returns full rank
>>>>> (10) in about 8 percent of cases (see appended script).
>>>>>
>>>>> I think this is a tolerance problem. The ``matrix_rank`` algorithm
>>>>> does this by default:
>>>>>
>>>>> S = spl.svd(M, compute_uv=False)
>>>>> tol = S.max() * np.finfo(S.dtype).eps
>>>>> return np.sum(S > tol)
>>>>>
>>>>> I guess we'd we want the lowest tolerance that nearly always or always
>>>>> identifies numerically rank deficient matrices. I suppose one way of
>>>>> looking at whether the tolerance is in the right range is to compare
>>>>> the calculated tolerance (``tol``) to the minimum singular value
>>>>> (``S.min()``) because S.min() in our case should be very small and
>>>>> indicate the rank deficiency. The mean value of tol / S.min() for the
>>>>> current algorithm, across many iterations, is about 2.8. We might
>>>>> hope this value would be higher than 1, but not much higher, otherwise
>>>>> we might be rejecting too many columns.
>>>>>
>>>>> Our current algorithm for tolerance is the same as the 2-norm of M *
>>>>> eps. We're citing Golub and Van Loan for this, but now I look at our
>>>>> copy (p 261, last para) - they seem to be suggesting using u * |M|
>>>>> where u = (p 61, section 2.4.2) eps / 2. (see [1]). I think the Golub
>>>>> and Van Loan suggestion corresponds to:
>>>>>
>>>>> tol = np.linalg.norm(M, np.inf) * np.finfo(M.dtype).eps / 2
>>>>>
>>>>> This tolerance gives full rank for these rank-deficient matrices in
>>>>> about 39 percent of cases (tol / S.min() ratio of 1.7)
>>>>>
>>>>> We see on p 56 (section 2.3.2) that:
>>>>>
>>>>> m, n = M.shape
>>>>> 1 / sqrt(n) . |M|_{inf} <= |M|_2
>>>>>
>>>>> So we can get an upper bound on |M|_{inf} with |M|_2 * sqrt(n). Setting:
>>>>>
>>>>> tol = S.max() * np.finfo(M.dtype).eps / 2 * np.sqrt(n)
>>>>>
>>>>> gives about 0.5 percent error (tol / S.min() of 4.4)
>>>>>
>>>>> Using the Mathworks threshold [2]:
>>>>>
>>>>> tol = S.max() * np.finfo(M.dtype).eps * max((m, n))
>>>>>
>>>>> There are no false negatives (0 percent rank 10), but tol / S.min() is
>>>>> around 110 - so conservative, in this case.
>>>>>
>>>>> So - summary - I'm worrying our current threshold is too small,
>>>>> letting through many rank-deficient matrices without detection. I may
>>>>> have misread Golub and Van Loan, but maybe we aren't doing what they
>>>>> suggest. Maybe what we could use is either the MATLAB threshold or
>>>>> something like:
>>>>>
>>>>> tol = S.max() * np.finfo(M.dtype).eps * np.sqrt(n)
>>>>>
>>>>> - so 2 * the upper bound for the inf norm = 2 * |M|_2 * sqrt(n) . This
>>>>> gives 0 percent misses and tol / S.min() of 8.7.
>>>>>
>>>>> What do y'all think?
>>>>>
>>>>> Best,
>>>>>
>>>>> Matthew
>>>>>
>>>>> [1]
>>>>> http://matthew-brett.github.com/pydagogue/floating_error.html#machine-epsilon
>>>>> [2] http://www.mathworks.com/help/techdoc/ref/rank.html
>>>>>
>>>>> Output from script:
>>>>>
>>>>> Percent undetected current: 9.8, tol / S.min(): 2.762
>>>>> Percent undetected inf norm: 39.1, tol / S.min(): 1.667
>>>>> Percent undetected upper bound inf norm: 0.5, tol / S.min(): 4.367
>>>>> Percent undetected upper bound inf norm * 2: 0.0, tol / S.min(): 8.734
>>>>> Percent undetected MATLAB: 0.0, tol / S.min(): 110.477
>>>>>
>>>>> <script>
>>>>> import numpy as np
>>>>> import scipy.linalg as npl
>>>>>
>>>>> M = 40
>>>>> N = 10
>>>>>
>>>>> def make_deficient():
>>>>> X = np.random.normal(size=(M, N))
>>>>> deficient_X = X.copy()
>>>>> if M > N: # Make a column deficient
>>>>> deficient_X[:, 0] = deficient_X[:, 1] + deficient_X[:, 2]
>>>>> else: # Make a row deficient
>>>>> deficient_X[0] = deficient_X[1] + deficient_X[2]
>>>>> return deficient_X
>>>>>
>>>>> matrices = []
>>>>> ranks = []
>>>>> ranks_inf = []
>>>>> ranks_ub_inf = []
>>>>> ranks_ub2_inf = []
>>>>> ranks_mlab = []
>>>>> tols = np.zeros((1000, 6))
>>>>> for i in range(1000):
>>>>> m = make_deficient()
>>>>> matrices.append(m)
>>>>> # The SVD tolerances
>>>>> S = npl.svd(m, compute_uv=False)
>>>>> S0 = S.max()
>>>>> # u in Golub and Van Loan == numpy eps / 2
>>>>> eps = np.finfo(m.dtype).eps
>>>>> u = eps / 2
>>>>> # Current numpy matrix_rank algorithm
>>>>> ranks.append(np.linalg.matrix_rank(m))
>>>>> # Which is the same as:
>>>>> tol_s0 = S0 * eps
>>>>> # ranks.append(np.linalg.matrix_rank(m, tol=tol_s0))
>>>>> # Golub and Van Loan suggestion
>>>>> tol_inf = npl.norm(m, np.inf) * u
>>>>> ranks_inf.append(np.linalg.matrix_rank(m, tol=tol_inf))
>>>>> # Upper bound of |X|_{inf}
>>>>> tol_ub_inf = tol_s0 * np.sqrt(N) / 2
>>>>> ranks_ub_inf.append(np.linalg.matrix_rank(m, tol=tol_ub_inf))
>>>>> # Times 2 fudge
>>>>> tol_ub2_inf = tol_s0 * np.sqrt(N)
>>>>> ranks_ub2_inf.append(np.linalg.matrix_rank(m, tol=tol_ub2_inf))
>>>>> # MATLAB algorithm
>>>>> tol_mlab = tol_s0 * max(m.shape)
>>>>> ranks_mlab.append(np.linalg.matrix_rank(m, tol=tol_mlab))
>>>>> # Collect tols
>>>>> tols[i] = tol_s0, tol_inf, tol_ub_inf, tol_ub2_inf, tol_mlab, S.min()
>>>>>
>>>>> rel_tols = tols / tols[:, -1][:, None]
>>>>>
>>>>> fmt = 'Percent undetected %s: %3.1f, tol / S.min(): %2.3f'
>>>>> max_rank = min(M, N)
>>>>> for name, ranks, mrt in zip(
>>>>> ('current', 'inf norm', 'upper bound inf norm',
>>>>> 'upper bound inf norm * 2', 'MATLAB'),
>>>>> (ranks, ranks_inf, ranks_ub_inf, ranks_ub2_inf, ranks_mlab),
>>>>> rel_tols.mean(axis=0)[:5]):
>>>>> pcnt = np.sum(np.array(ranks) == max_rank) / 1000. * 100
>>>>> print fmt % (name, pcnt, mrt)
>>>>> </script>
>>>>
>>>>
>>>> The polynomial fitting uses eps times the largest array dimension for the
>>>> relative condition number. IIRC, that choice traces back to numerical
>>>> recipes.
>>
>> Chuck - sorry - I didn't understand what you were saying, and now I
>> think you were proposing the MATLAB algorithm. I can't find that in
>> Numerical Recipes - can you? It would be helpful as a reference.
>>
>>> This is the same as Matlab, right?
>>
>> Yes, I believe so, i.e:
>>
>> tol = S.max() * np.finfo(M.dtype).eps * max((m, n))
>>
>> from my original email.
>>
>>> If the Matlab condition is the most conservative, then it seems like a
>>> reasonable choice -- conservative is good so long as your false
>>> positive rate doesn't become to high, and presumably Matlab has enough
>>> user experience to know whether the false positive rate is too high.
>>
>> Are we agreeing to go for the Matlab algorithm?
>
> As extra data, current Numerical Recipes (2007, p 67) appears to prefer:
>
> tol = S.max() * np.finfo(M.dtype).eps / 2. * np.sqrt(m + n + 1.)
To add extra confusing flames to this fire, a survey of random
matrices of sizes M = (3, 5, 10, 50, 100, 500), N = (3, 5, 10, 50,
100, 500) in all combinations suggests that this last Numerical
Recipes algorithm does not give false negatives, and the treshold is
generally considerably lower than the MATLAB algorithm. At least for
this platform (linux 64 bit), and with only one rank deficient column
/ row.
My feeling is still that the risk of using the matlab version is less,
and the risk of too many false positives is relatively small. If
anyone disagrees, it might be worth running the test rig for other
parameters and platforms,
Best,
Matthew
import numpy as np
import numpy.linalg as npl
algs = (('current', lambda M, S, eps2: S.max() * eps2),
('inf norm', lambda M, S, eps2: npl.norm(M, np.inf) * eps2 / 2),
('ub inf norm',
lambda M, S, eps2: S.max() * eps2 / 2 * np.sqrt(M.shape[1])),
('ub inf norm * 2',
lambda M, S, eps2: S.max() * eps2 * np.sqrt(M.shape[1])),
('NR',
lambda M, S, eps2: S.max() * eps2 / 2 * np.sqrt(sum(M.shape + (1,)))),
('MATLAB', lambda M, S, eps2: S.max() * eps2 * max(M.shape)),
)
def make_deficient(M, N, loc=0, scale=1):
X = np.random.normal(size=(M, N))
deficient_X = X.copy()
if M > N: # Make a column deficient
deficient_X[:, 0] = deficient_X[:, 1] + deficient_X[:, 2]
else: # Make a row deficient
deficient_X[0] = deficient_X[1] + deficient_X[2]
return deficient_X
def doit(algs=algs):
results = {}
n_iters = 1000
n_algs = len(algs)
pcnt_div = n_iters * 100.
tols = np.zeros((n_iters, n_algs))
ranks = np.zeros((n_iters, n_algs))
eps2 = np.finfo(float).eps
for M in (3, 5, 10, 50, 100, 500):
for N in (3, 5, 10, 50, 100, 500):
max_rank = min(M, N)
svs = np.zeros((n_iters, max_rank))
for loc in (0, 100, 1000):
for scale in (1, 100, 1000, 10000):
for i in range(n_iters):
m = make_deficient(M, N, loc, scale)
# The SVD tolerances
S = npl.svd(m, compute_uv=False)
svs[i] = np.sort(S)
for j, alg in enumerate(algs):
name, func = alg
tols[i, j] = func(m, S, eps2)
ranks[i, j] = np.sum(S > tols[i, j])
del m, S
rel_tols = tols / svs[:, 0][:, None]
key = (M, N, loc, scale)
print key
pcnts = np.sum(ranks == max_rank, axis=0) / pcnt_div
mrtols = np.mean(rel_tols, axis=0)
results[key] = (pcnts, mrtols)
More information about the NumPy-Discussion
mailing list