def GP(data, init_haps, positions, N, r, times, numThreads=1, pool=None):
if numThreads==1:
results=[GP.singleLocus(data[:, (i,), :],init_haps[:, (i,)], N, r, times, pos) for i, pos in enumerate(positions)]
else:
params=[{'X':data[:, (i,), :],'initHaps':init_haps[:, (i,)], 'N':N, 'r': r, 'times':times, 'pos':pos } for i, pos in enumerate(positions)]
results=pool.map(gpHelper,params)
df=pd.DataFrame(map(lambda x:x.values(),results),columns=results[0].keys()).astype(float);df.index=df.pos.astype(int);df.drop('pos',axis=1,inplace=True)
return df
def runGPTimes():
import popgen.TimeSeries.GaussianProcess.Estimate as GP
reload(GP)
sim=Simulation.Simulation.Load(s=0.1)
times=[]
for nloci in [1, 3, 5, 7, 10]:
for _ in range(200):
l=np.sort(np.random.choice(sim.X.shape[1],nloci,replace=False))
times+=[(nloci,GP.multipleLocus(sim.X[:, l, :],sim.H0.values[:,l], sim.N, sim.r, sim.getGenerationTimes(), np.array(sim.positions)[l]).Time, sim.X[0,l[0]].mean())]
print times[-1]
df=pd.DataFrame(times,columns=['n','time','x0'])
df.to_pickle(utl.outpath+'ROC/GPTimes.df')
def gpHelper(args):
return GP.singleLocus(**args)
