def _estimate(self, X, var, n_factors):
n, p = X.numRows(), X.numCols()
old_ll = -numpy.inf
llconst = p * numpy.log(2. * numpy.pi) + n_factors
psi = numpy.ones(p, dtype=numpy.float32)
nsqrt = numpy.sqrt(n)
logliks = []
logger.info("Computing factor analysis")
for _ in range(self.max_iter):
sqrt_psi = numpy.sqrt(psi) + self.__eps
s, V, unexp_var = svd(self._tilde(X, sqrt_psi, nsqrt), n_factors)
s = s**2
W = self._update_factors(s, V, sqrt_psi)
ll = self._loglik(llconst, unexp_var, s, psi, n)
logliks.append(ll)
psi = self._update_variance(var, W)
if abs(ll - old_ll) < self.threshold:
break
old_ll = ll
return W, logliks, psi
def _compute_pcs(X):
sds, loadings, _ = svd(X)
sds = sds / scipy.sqrt(max(1, X.numRows() - 1))
return loadings, sds
def _whiten(self, X):
s, v, _ = svd(X, X.numCols())
K = (v.T / s)[:, :self.n_components]
S = K * scipy.sqrt(X.numRows())
S = DenseMatrix(S.shape[0], S.shape[1], S.flatten(), True)
return X.multiply(S), K