davemcg
commented
6 years ago I'm having the same issue. Is msCentipede not being updated anymore?
Open xubeisi opened 7 years ago
davemcg
commented
6 years ago I'm having the same issue. Is msCentipede not being updated anymore?
msCentipede works fine for test data, but when I tried applied to other data with ATAC-seq model, learn model gives
/home/bxu2/opt/msCentipede/mscentipede.py:15: RuntimeWarning: overflow encountered in exp logistic = lambda x: 1./(1+np.exp(x)) /home/bxu2/opt/msCentipede/mscentipede.py:598: RuntimeWarning: overflow encountered in exp func = arg * zeta.estim[:,1:] - nplog(1 + np.exp(arg)) Restart 1 ... Traceback (most recent call last): File "/home/bxu2/opt/msCentipede/msCentipede_call.py", line 279, in
main()
File "/home/bxu2/opt/msCentipede/msCentipede_call.py", line 272, in main
learn_model(options)
File "/home/bxu2/opt/msCentipede/msCentipede_call.py", line 47, in learn_model
background_counts, options.model, options.restarts, options.mintol)
File "/home/bxu2/opt/msCentipede/mscentipede.py", line 1133, in estimate_optimal_model
alpha, beta, omega, pi_null, tau_null, model)
File "/home/bxu2/opt/msCentipede/mscentipede.py", line 982, in square_EM
EM(data, scores, zeta, pi, tau, alpha, beta, omega, pi_null, tau_null, model)
File "/home/bxu2/opt/msCentipede/mscentipede.py", line 926, in EM
beta.update(scores, zeta)
File "/home/bxu2/opt/msCentipede/mscentipede.py", line 628, in update
self.estim = optimizer(xo, function, gradient, hessian, scores=scores, zeta=zeta)
File "/home/bxu2/opt/msCentipede/mscentipede.py", line 704, in optimizer
solution = solvers.cp(F)
File "build/bdist.linux-x86_64/egg/cvxopt/cvxprog.py", line 1969, in cp
File "build/bdist.linux-x86_64/egg/cvxopt/cvxprog.py", line 715, in cpl
OverflowError: (34, 'Numerical result out of range')
I googled found this is something seems related to the precision limitation, but I am wondering is this something I can correct in msCentipede or I have to dig into cvxopt?
Thanks!