def rundist_2loaders(arg):
l1, l2, nc = arg
m = Data().fit(l1(),
l2(),
debug_ftsel=False,
quiet=True,
pca=20,
titles=("3", "6"),
make_even=True)
labels = p.gmm_2(*m.dx, nc=nc, cov='full')
labels2 = p.gmm_2(*m.dx, nc=10, cov='full')
a, b = Q.rari_score(*labels, *m.dx)
c, d = Q.rari_score(*labels2, *m.dx)
return a, c
def samplenum(arg):
loader, samp = arg
t = time.time()
m = Data().fit(*loader(seed=None, subsample=samp),
debug_ftsel=False,
quiet=True,
pca=20,
titles=("3", "6"),
make_even=True)
#labels = p.gmm_2(*m.dx,nc=nc,cov='full')
#a,b = Q.rari_score(*labels, *m.dx)
t1 = time.time() - t
a = Q.rari_score(*p.gmm_2(*m.dx, nc=15, cov='full'), *m.dx)[0]
t2 = time.time() - t
b = Q.rari_score(*p.gmm_2(*m.dx, nc=15, cov='tied'), *m.dx)[0]
return a, b, t1, t2
def rundist(arg):
loader, nc, scale = arg
m = Data().fit(*loader(),
debug_ftsel=False,
scale=scale,
maxgenes=800,
quiet=True,
pca=20,
titles=("3", "6"),
make_even=True)
clust = lambda ncc: Q.rari_score(*p.gmm_2(*m.dx, nc=ncc, cov='full'), *m.dx
)[0]
return [clust(ncc) for ncc in nc]
def get_noise_run(args):
loader, pool, cluster = args
adat = loader()
noiserange = range(0, 110, 10)
if type(adat.X) != csr_matrix:
adat.X = csr_matrix(adat.X)
# i could use mp in get_noise_data i think
rdydata = noise.get_noise_data(adat,
noiserange,
title="magic",
poolsize=pool,
cluster=cluster)
r = [Q.rari_score(*m.labels, *m.dx) for m in rdydata]
return r
def normal(arg):
l1, l2, pca, scale = arg
m = Data().fit(l1(),
l2(),
debug_ftsel=False,
quiet=True,
scale=scale,
pca=pca,
dimensions=20,
titles=("3", "6"),
make_even=True)
#labels = p.gmm_2(*m.dx,nc=nc,cov='full')
#a,b = Q.rari_score(*labels, *m.dx)
nc = [15]
return [
Q.rari_score(*p.gmm_2(*m.dx, nc=NC, cov='full'), *m.dx)[0] for NC in nc
]
def get_noise_run_moar(args):
loader, cluster, level, metrics = args
if level == 0:
return [1] * len(cluster) * len(
metrics) # should be as long a a normal return
adat = loader()
if type(adat.X) != csr_matrix:
adat.X = csr_matrix(adat.X)
# todo: also loop over cluster algos
m = noise.get_noise_single(adat, level)
r = [
Q.rari_score(*c(*m.dx), *m.dx, metric=metric) for c in cluster
for metric in metrics
]
return r
dimensions=umapdim,
ft_combine = lambda x,y: x or y,
umap_n_neighbors=10, # used in example
maxmean= 4,
minmean = 0.02,
corrcoef=False,
mitochondria = "mt-",
pp='linear',
debug_ftsel=False,
scale=False,
quiet = True,
titles = ("3","6"),
make_even=True)
### get numberz :)
getnumber = lambda y,x: q.rari_score(*y,*x)[0]
def clustandscore(d):
t = h.cluster_ab(*d.dx, nc =9, cov = 'tied' )
f = h.cluster_ab(*d.dx, nc =9, cov = 'full' )
#print ("asdasd",t,f,d.dx)
t= getnumber(t,d.dx)
f= getnumber(f,d.dx)
#print(f"{t} {f}")
return t,f
def get_data_mp(x):
return clustandscore(getdata(*x))