def tosimilaritypvalues(self, axis, metric='cosine'):
if axis == 0:
return datamatrix(rowname=self.rowname,
rowlabels=self.rowlabels.copy(),
columnname=self.rowname,
columnlabels=self.rowlabels.copy(),
matrixname=self.rowname + '-' + self.rowname +
'_' + metric + '_similarity_derived_from_' +
self.matrixname,
matrix=mlstats.corr(self.matrix,
axis=axis,
metric=metric,
getpvalues=True)[1],
rowmeta=copy.deepcopy(self.rowmeta),
columnmeta=copy.deepcopy(self.rowmeta))
elif axis == 1:
return datamatrix(rowname=self.columnname,
rowlabels=self.columnlabels.copy(),
columnname=self.columnname,
columnlabels=self.columnlabels.copy(),
matrixname=self.columnname + '-' +
self.columnname + '_' + metric +
'_similarity_derived_from_' + self.matrixname,
matrix=mlstats.corr(self.matrix,
axis=axis,
metric=metric,
getpvalues=True)[1],
rowmeta=copy.deepcopy(self.columnmeta),
columnmeta=copy.deepcopy(self.columnmeta))
else:
raise ValueError('invalid axis')
# single bead move
new_x0 = old_x0
new_x1 = old_x1 + np.random.randn() * opt_sig(omega, tau, lam)
path[islice] = new_x0
path[(islice + 1) % nslice] = new_x1
# calculate action change
new_action = action(path, omega, beta, lam)
prob = np.exp(-(new_action - old_action)) # action = -ln_rho
# accept or recject
if np.random.rand() < prob:
naccept += 1
else:
path[islice] = old_x0
path[(islice + 1) % nslice] = old_x1
# end if prob
nmove += 1
# end for islice
# end for istep
# end for isample
np.savetxt('pos.dat', pos)
pos2 = pos[nequil:] * pos[nequil:]
x2eq = x2[nequil:]
accept = float(naccept) / nmove #float(naccept)/nsample/nslice/block_size
print(omega, beta, sig, accept, corr(pos2), np.mean(x2eq), error(x2eq))
# end __main__
return args.omega, args.beta, args.lam, args.sig, args.nequil, args.nsample, args.use_opt_sig, args.block_size
# end def
if __name__ == '__main__':
omega, beta, lam, sig, nequil, nsample, use_opt_sig, block_size = parse_inputs(
)
# calculate optimal step size
if use_opt_sig:
sig = opt_sig(omega, beta, lam)
# end if
accept, x2 = sample_1d_harmonic_analytic(ln_rho, omega, beta, lam, sig,
block_size)
x2eq = x2[nequil:]
fmt = "{omega:6.2f} {beta:4.2f} {sig:4.2f} {accept:3.2f} {corr:3.1f} {x2:4.6f} {x2e:4.6f}"
output = fmt.format(omega=omega,
beta=beta,
sig=sig,
accept=accept,
corr=corr(x2eq),
x2=np.mean(x2eq),
x2e=error(x2eq))
print(output)
#print( omega,beta,sig,accept,corr(x2eq),np.mean(x2eq),error(x2eq) )
np.savetxt('ax2.dat', x2)
# end __main__
def least_squares_fit(x, y):
"""x와 y가 학습 데이터로 주어졌을 때
오류의 제곱 값을 최소화해 주는 알파와 베타를 계산"""
beta = corr(x, y) * stdev(y) / stdev(x)
alpha = mean(y) - beta * mean(x)
return alpha, beta