def gen_predictions():
#--setup nlo code
set_mstw()
#--setup pdfs
hq.setup(fname.ljust(100), 0)
nsets = 97
#--read kinematics
t = pd.read_excel('expdata/eic.xlsx')
t = t.to_dict(orient='list')
npts = len(t['X'])
#--compute F2c accross kinematics
for iset in range(nsets):
hq.setup_pdfset(iset)
t[iset] = [] #--create empty array to fill theory
for i in range(npts):
x = t['X'][i]
Q2 = t['Q2'][i]
Q = np.sqrt(Q2)
#,f2,f2c,f2b,fl,flc,flb
F2c = hq.mstwnc(x, Q, 1)[1]
t[iset].append(F2c)
#--convert all arrays to numpy arrays
for _ in t:
t[_] = np.array(t[_])
checkdir('data')
save(t, 'data/predictions.po')
def get_vocab_info(doc, train_idx, min_freq, max_vocab_size, n_gram, dir_path):
if os.path.exists(os.path.join(dir_path, settings.vocab_file)):
vocab_info = tools.load(os.path.join(dir_path, settings.vocab_file))
if vocab_info['min_freq'] == min_freq and vocab_info['max_vocab_size'] == max_vocab_size \
and vocab_info['df'] == len(train_idx) and vocab_info['n_gram'] == n_gram:
return vocab_info
tf = Counter()
data_doc = [doc[i] for i in train_idx]
for i, doc in enumerate(data_doc):
for tf_tuple in doc:
term, frequency = tf_tuple
tf[term] += frequency
tf_tuples = sorted(tf.items(), key=lambda x:
(-x[1], x[0])) # sort by frequency, then alphabetically
stoi = dict()
itos = dict()
valid_term = 0
for term, freq in tf_tuples:
if freq >= min_freq and valid_term < max_vocab_size:
stoi[term] = valid_term
itos[valid_term] = term
valid_term += 1
else:
tf.pop(term)
vocab_info = {
"min_freq": min_freq,
"max_vocab_size": max_vocab_size,
'n_gram': n_gram,
"stoi": stoi,
"itos": itos,
"tf": tf,
'df': len(train_idx)
}
tools.save(os.path.join(dir_path, settings.vocab_file), vocab_info)
return vocab_info
def main():
"""
Main program loop.
"""
args = setup_args()
"""
Get the meta arguments.
"""
cmd = tools.meta(args)
args = vars(args)
corpus = load_corpus(filename=args.pop('file'), clean=args.pop('clean'))
detector = create_detector(model=args.pop('model'),
extractor=args.pop('extractor'),
scorer=args.pop('scorer'),
filter=args.pop('filter'),
corpus=corpus,
**args)
resolved, extrapolated = detect(detector=detector, corpus=corpus)
cmd['model'] = str(type(detector).__name__)
cmd['extractor'] = str(type(detector.extractor).__name__)
cmd['scorer'] = str(type(detector.scorer).__name__)
cmd['filter'] = str(type(detector.filter).__name__)
cmd['resolver'] = str(type(detector.resolver).__name__)
cmd['extraploator'] = str(type(detector.extrapolator).__name__)
cmd['postprocessor'] = str(type(detector.postprocessor).__name__)
tools.save(args['output'], {
'meta': cmd,
'resolved': resolved,
'extrapolated': extrapolated
})
def runEndless(self, filename, interval=15):
"""Run tournaments endlessly until terminated, and save data every so often. The interval between
saves/progress reports can be modifed by the keyword interval (minutes)."""
start = time.time()
last_report = start
t = 0 # Counter for triggering generational recording of statistics
while True: # Evolutionary loop
self.tournament() # Run one tournament
if (t % self.popsize == 0): # Record statistics every generation
af, bf, d, c = self.fitStats()
self.avgHistory.append(af)
self.bestHistory.append(bf)
self.divHistory.append(d)
self.conHistory.append(c)
self.dateEdited = str(date.today())
if (time.time() - last_report) > (
interval * 60
): # If it's been more than N minutes since last report/save, save/report
print("\nSaving...")
save(filename, self)
print('Generations Run: %i' %
int(self.generationsRun)) # Print generations run so far
print('Fitness Avg=%f, Max=%f, Div=%f, Con=%f' %
tuple(self.fitStats()))
print('Time elapsed: %f sec / %f min / %f hours' %
((time.time() - start), (time.time() - start) / 60,
(time.time() - start) / 3600))
last_report = time.time()
t += 1
async def on_ready():
"""Preliminary activity
Checks all guilds for all users and initializes them in memory.
"""
print('We have logged in as {0.user}.'.format(client))
print('I will be speaking "{0}" as default.'.format(language))
print("Looking for new members/guilds while I was away...")
# We search all guilds and users Milton can see and initialize them.
i = 0
for guild in client.guilds:
# Add new guilds
if str(guild.id) not in G.GLD.keys():
default_dict = {"language": language}
G.GLD[str(guild.id)] = mun.DefaultMunch().fromDict(default_dict, 0)
tools.save(G.OPT.guilds_path, G.GLD)
# Add new members
for member in guild.members:
if str(member.id) not in G.USR.keys():
i += 1
G.USR[str(member.id)] = mun.DefaultMunch(0)
G.USR[str(member.id)].name = member.name
tools.save_users()
print(f"Found {i} new members")
game = ds.Game("with myself.") # Update "playing" message, for fun.
print("Ready!")
await client.change_presence(status=ds.Status.online, activity=game)
def get_vocab_info(doc, labels, train_idx, output_path, sparse_format=False):
if os.path.exists(os.path.join(output_path, settings.vocab_file)):
vocab_info = tools.load(os.path.join(output_path, settings.vocab_file))
if len(vocab_info['vocab_dict']) <= settings.max_vocab_size:
return vocab_info
tf = Counter()
data_doc = [doc[i] for i in train_idx]
leaf_label = labels[-1][train_idx]
for i, x in enumerate(data_doc):
for word_tuple in x:
word, frequency = word_tuple
if sparse_format or (word not in stop_words
and not word.isnumeric()):
tf[word] += frequency
vocab_dict = dict()
new_tf = Counter()
for i, v in enumerate(tf.most_common(settings.max_vocab_size)):
vocab_dict[v[0]] = i
new_tf[v[0]] = tf[v[0]]
tf = new_tf
tf["<DF>"] = len(data_doc) # to store the number of documents
vocab_info = {"vocab_dict": vocab_dict, "tf": tf}
tools.save(os.path.join(output_path, settings.vocab_file), vocab_info)
return vocab_info
def gen_hess_error():
t = load('data/predictions.po')
nsets = 97
err2 = np.zeros(t[0].size)
for iset in range(1, nsets, 2):
err2 += (t[iset] - t[iset + 1])**2 / 4.0
t['hess'] = err2**0.5
save(t, 'data/hess.po')
def split_train_test(data, classes, rate, output_dir, seed=0):
train_idx, test_idx = split(data, classes, rate, seed=seed)
tools.make_sure_path_exists(output_dir)
tools.save(os.path.join(output_dir, 'train_test_idx.npz'), {
'train_idx': train_idx,
'test_idx': test_idx
})
return train_idx, test_idx
def start(self):
best_fit = settings.min_fit
best_index = -1
for i in range(self.size):
draw = Draw()
draw.set_fit(self.target)
self.population[i] = draw
if (draw.fit <= best_fit):
best_index = i
best_fit = draw.fit
tools.save(self.population[best_index], best_fit, 0)
def gen_rand(nrep=1000):
"""
random pdfs will be constructed from eigen directions as
f_k = f_0 + sum_i rand_i (f[i]-f[-i])/2
similarly the random F2c will computed as
F2c_k = F2c_0 + sum_i rand_i (F2c[i]-F2c[-i])/2
here we generate the rand_i
"""
nsets = 97
rand = np.random.randn(nrep, (nsets - 1) / 2)
save(rand, 'data/rand.po')
def split_label_unlabel(data, index, classes, rate, output_dir, seed=0):
label_idx, unlabel_idx = split_data.split(data,
classes,
rate,
index=index,
seed=seed)
tools.make_sure_path_exists(output_dir)
tools.save(os.path.join(output_dir, 'label_unlabel_idx.npz'), {
'label_idx': label_idx,
'unlabel_idx': unlabel_idx
})
return label_idx, unlabel_idx
async def on_guild_join(guild):
"""Handles when Milton joins a guild"""
if str(guild.id) not in G.GLD.keys():
default_dict = {"language": language}
G.GLD[str(guild.id)] = mun.DefaultMunch().fromDict(default_dict, 0)
tools.save(G.OPT.guilds_path, G.GLD)
# Add new members
for member in guild.members:
if str(member.id) not in G.USR.keys():
G.USR[str(member.id)] = mun.DefaultMunch(0)
G.USR[str(member.id)].name = member.name
tools.save_users()
def get_old_hessian(self):
F=open('data/correlation_matrix.dat')
L=F.readlines()
F.close()
L=[l.strip().split() for l in L if l.strip()!='']
L=[[float(x) for x in l] for l in L]
corr=np.array(L)
ndim=len(L)
cov=np.zeros((ndim,ndim))
for i in range(ndim):
for j in range(ndim):
cov[i,j]=corr[i,j]*self.dP0[i]*self.dP0[j]
hess=LA.inv(cov)
save(hess,'data/old.hess')
def change_language_logic(message):
prefix_length = len((G.OPT.prefix + G.LOC.commands.changeLang.id)) + 1
query = message.content[prefix_length:(prefix_length + 2)].lower()
available_locales = G.GLOC.keys()
out = tools.MsgBuilder()
if query in available_locales:
G.GLD[str(message.guild.id)].language = query
tools.save(G.OPT.guilds_path, G.GLD)
G.update_loc(G.GLOC[G.GLD[str(message.guild.id)].language])
out.add(G.LOC.commands.changeLang.success.format(query.upper()))
return out.parse()
else:
locales = " ".join(available_locales).upper()
out.add(G.LOC.commands.changeLang.error.format(query, locales))
return out.parse()
async def saveall():
await say(systemchannel, 'saving all dicts')
print('saving all dicts')
tools.save(mods, 'mods.json')
tools.save(localdict, 'localdict.json')
tools.save(authinfo, 'authinfo.json')
tools.save(commands, 'commands.json')
def gen_hess_glue(Q2=2.0):
X = 10**np.linspace(-4, -1, 100)
X = np.append(X, np.linspace(0.1, 0.99, 100))
data = {'X': X, 'Q2': Q2}
pdfs = lhapdf.mkPDFs(fname)
for i in range(len(pdfs)):
data[i] = np.array([pdfs[i].xfxQ2(21, x, Q2) for x in X])
err2 = np.zeros(data[0].size)
for iset in range(1, len(pdfs), 2):
err2 += (data[iset] - data[iset + 1])**2 / 4.0
data['err'] = err2**0.5
save(data, 'data/hess_glue_%.2f.po' % Q2)
def encode_chat_text_to_vectors(patterns):
data = tools.load_if_exists("tmp.pickle")
if data is not None:
doc, doc_vecs, reponse_patterns = data
else:
doc = list()
doc_vecs = list()
reponse_patterns = list()
print("encoding sentences")
for i,pattern in tqdm(enumerate(patterns), total=len(patterns)):
doc.extend(pattern.input)
vectors = [encode(line) for line in pattern.input]
doc_vecs.extend(vectors)
reponse_patterns.extend([pattern]*len(pattern.input))
tools.save("tmp.pickle",[doc,doc_vecs,reponse_patterns])
return doc, np.array(doc_vecs), reponse_patterns
def main():
"""
Main program loop.
"""
args = setup_args()
tfidf = construct(
file=args.file,
remove_retweets=args.remove_retweets,
normalize_words=args.normalize_words,
character_normalization_count=args.character_normalization_count,
remove_unicode_entities=args.remove_unicode_entities,
stem=args.stem)
cmd = tools.meta(args)
pcmd = tools.meta(args)
tools.save(args.output, {'cmd': cmd, 'pcmd': pcmd, 'tfidf': tfidf})
def gen_mc_F2c():
rnd = load('data/rand.po')
t = load('data/predictions.po')
#--gen MCF2c
nsets = 97
mcF2c = []
for k in range(len(rnd)):
F2c = np.copy(t[0])
cnt = 0
for iset in range(1, nsets, 2):
F2c += rnd[k][cnt] * (t[iset] - t[iset + 1]) / 2
cnt += 1
mcF2c.append(F2c)
mcF2c = np.array(mcF2c)
save(mcF2c, 'data/mcF2c.po')
def gen_mc_glue(Q2=2.0):
rnd = load('data/rand.po')
t = load('data/hess_glue_%.2f.po' % Q2)
#--gen MCF2c
nsets = 97
mcglue = []
for k in range(len(rnd)):
g = np.copy(t[0])
cnt = 0
for iset in range(1, nsets, 2):
g += rnd[k][cnt] * (t[iset] - t[iset + 1]) / 2
cnt += 1
mcglue.append(g)
mcglue = np.array(mcglue)
save(mcglue, 'data/mcglue_%.2f.po' % Q2)
def run_continue(filename, generations, save_interval):
"""Run a set number of tournaments, saving along the way every save_interval generations."""
start = time.time()
mga = read(filename)
for g in range(int(
generations /
save_interval)): # Run X generations and save every Y generations
print('Running generations %i - %i...' % (g * save_interval,
(g + 1) * save_interval))
# Run simulation
mga.runTournaments(save_interval * mga.popsize, report=True)
# Save data
generation = int(mga.generationsRun)
date = mga.dateEdited
filename = '%s_G%i_%s' % (filename[:-14], generation, date)
save(filename, mga)
print('%f sec elapsed so far \n' % (time.time() - start))
def _X_get_scan_data(self,fname):
F=open(fname)
L=F.readlines()
F.close()
L=[l.split() for l in L]
# construct headres
H={}
cnt=0
for h in L[0][2:]:
H[h]=cnt
cnt+=1
print H
# remove header table
L=L[1:]
# list or parameters
K=['BLNY','Nu_c','Nd_c','au_c','ad_c','bu_c','bd_c','Nu_T','Nd_T','au_T','ad_T','bu_T','bd_T']
# data to dict
D={}
for l in L:
D[l[0]]=[]
for l in L:
D[l[0]].append([float(x) for x in l[2:]])
data={}
chi20=D['FIT'][0][0]
for k in K:
data[k]={'h':[],'dchi2dxdx':[]}
pars=D[k]
i=H[k]
for j in range(0,len(pars),2):
chi2max=pars[j][0]
chi2min=pars[j+1][0]
pmax=pars[j][i]
pmin=pars[j+1][i]
h=0.5*(pmax-pmin)
dchi2dxdx=(chi2max-2*chi20+chi2min)/h**2
data[k]['h'].append(h)
data[k]['dchi2dxdx'].append(dchi2dxdx)
save(data,'data/%s.dchid2dxdx'%fname.split('/')[1].replace('.dat',''))
def crossover(pop):
best_fit = settings.min_fit
best_index = -1
new_pop = Population(pop.size, pop.target)
new_pop.gen = pop.gen + 1
for i in range(pop.size):
parent_a = pop.__select_parent(pop)
parent_b = pop.__select_parent(pop)
child = Draw.crossover(parent_a, parent_b)
child.mutate()
child.set_fit(pop.target)
if (child.fit <= best_fit):
best_index = i
best_fit = child.fit
new_pop.population[i] = child
tools.save(new_pop.population[best_index], best_fit, new_pop.gen)
print best_fit
return new_pop
def generate_dataset(n_games, skill=20, threads=1):
def play(engine, limit=0.01):
board = chess.Board()
# raw_history = []
history = []
while not board.is_game_over():
result = engine.play(board, chess.engine.Limit(time=limit))
# raw_history.append(board.copy())
history.append(to_bitboard(board, pack=True))
board.push(result.move)
return history, board.result()
def create_empty_games():
return {WHITE_WON: [], BLACK_WON: []}
engine = chess.engine.SimpleEngine.popen_uci(engine_path, timeout=None)
engine.configure({"Threads": threads, "Skill Level": skill})
games = create_empty_games()
print('playing....')
i = 1
while i != n_games + 1:
history, result = play(engine)
if result != DRAW:
print(f'.....{i}/{n_games}')
games[result].append((history, results[result]))
if i % 1000 == 0:
filename = f'dataset-{i // 1000}'
date = get_date()
save(games[WHITE_WON],
f'{filename}-white-{len(games[WHITE_WON])}-{date}')
save(games[BLACK_WON],
f'{filename}-black-{len(games[BLACK_WON])}-{date}')
games = create_empty_games()
i += 1
engine.quit()
def start_evolution():
conta = 0
generation = 0
sig = 0
draw = Draw()
error = Fitness.fitness(draw,target_matrix)[0]
while(1):
conta = conta + 1
new_draw = copy.deepcopy(draw).mutate()
if (new_draw.is_dirty()):
generation = generation + 1
print conta,generation,sig
new_error,new_img = Fitness.fitness(new_draw,target_matrix)
print new_error,"=erro novo /",error,"=erro antigo"
if (new_error <= error):
sig = sig + 1
draw = new_draw
draw.dirty = False
Tools.save(new_img,new_error,generation)
error = new_error
def get_data(self) -> dict or None:
timeout = eventlet.Timeout(10)
try:
response = requests.get(
self.URL_VK,
params={
"domain": self.DOMAIN,
"count": "5", # <100
"access_token": self.TOKEN,
"v": self.VERSION,
},
proxies={"https": "188.191.165.92:8080"})
data = response.json()
logging.info(data)
logging.info("finish scanning")
save(data)
return data
except eventlet.timeout.Timeout:
logging.warning(
'Got Timeout while retrieving VK JSON data. Cancelling...')
return None
finally:
timeout.cancel()
def gen_weights():
rnd = load('data/rand.po')
t = load('data/predictions.po')
mcF2c = load('data/mcF2c.po')
#--get absolute simulated errors
t['alpha'] = t[0] * t['relerr']
#--gen chi2
chi2 = []
for F2c in mcF2c:
exp = np.copy(t[0])
res = (exp - F2c) / t['alpha']
chi2.append(np.sum(res**2))
chi2 = np.array(chi2)
dchi2 = chi2 - t[0].size
#--gen weights
weights = np.exp(-0.5 * dchi2)
norm = np.sum(weights)
weights /= norm
save(weights, 'data/weights.po')
def ProcessPIV(args, bga, bgb, reflection, stg):
# read images into numpy arrays
file_a, file_b, counter = args
frame_a = tools.imread(file_a)
frame_b = tools.imread(file_b)
# removing background and reflections
if bga is not None:
frame_a = frame_a - bga
frame_b = frame_b - bgb
frame_a[reflection == 255] = 0
frame_b[reflection == 255] = 0
#applying a static mask (taking out the regions where we have walls)
pnts = draw.polygon(stg['YM'], stg['XM'], frame_a.shape)
frame_a[pnts] = 0
frame_b[pnts] = 0
plt.imshow(frame_a, cmap='gray')
plt.show()
# main piv processing
u, v = pyprocess.extended_search_area_piv( frame_a, frame_b, \
window_size=stg['WS'], overlap=stg['OL'], dt=stg['DT'], search_area_size=stg['SA'], sig2noise_method=None)
x, y = pyprocess.get_coordinates(image_size=frame_a.shape,
window_size=stg['WS'],
overlap=stg['OL'])
u, v, mask = validation.local_median_val(u, v, 2000, 2000, size=2)
if stg['BVR'] == 'on':
u, v = filters.replace_outliers(u,
v,
method='localmean',
max_iter=10,
kernel_size=2)
u, *_ = smoothn(u, s=0.5)
v, *_ = smoothn(v, s=0.5)
x, y, u, v = scaling.uniform(x, y, u, v, stg['SC'])
# saving the results
save_file = tools.create_path(file_a, 'Analysis')
tools.save(x, y, u, v, mask, save_file + '.dat')
def _get_scan_data(self,fname):
F=open(fname)
L=F.readlines()
F.close()
L=[l.split() for l in L]
# get parameter names
params=sorted(set([l[0] for l in L[2:]]))
# get parameters column index
icols={}
for p in params:
for i in range(len(L[0])):
if L[0][i]==p: icols[p]=i
# get dchidxdx
data={}
chi20=float(L[1][2])
for p in params:
data[p]={'h':[],'dchi2dxdx':[]}
rows=[l for l in L if l[0]==p]
icol=icols[p]
for i in range(0,len(rows),2):
chi2max=float(rows[i][2])
chi2min=float(rows[i+1][2])
pmax=float(rows[i][icol])
pmin=float(rows[i+1][icol])
h=0.5*(pmax-pmin)
dchi2dxdx=(chi2max-2*chi20+chi2min)/h**2
data[p]['h'].append(h)
data[p]['dchi2dxdx'].append(dchi2dxdx)
fmt='%s %0.4e %0.4e %0.2e %0.2e %0.2e %0.2e %0.4e'
print fmt%(p,pmin,pmax,chi20,chi2min,chi2max,h,dchi2dxdx)
save(data,'data/%s.dchid2dxdx'%fname.split('/')[1].replace('.dat',''))
def plotter(train = None, title = "", ylabel = "", fn = "test", plot = True, add_chpts = False, add_cv = True):
#figure
fig = plt.figure(facecolor='w', figsize=(12, 7))
ax = fig.add_subplot(111)
plt.title(title, fontsize=18, y = 1.05)
### train result
train_fig = train.model.plot(train.forecast, ax = ax, xlabel = "Date", ylabel = ylabel)
if add_chpts:
cpts = add_changepoints_to_plot(train_fig.gca(), train.model, train.forecast)
ax = plt.gca(); ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
c_txt = "RMSE=%.1f MAPE=%.1f%%"%(train.param.rmse, train.cv_metrics.manual_mape)
#format date
myFmt = mdates.DateFormatter('%d-%m')
ax.xaxis.set_major_formatter(myFmt)
#ticks - locator puts ticks at regular intervals
xloc = plticker.MultipleLocator(base=3.0)
ax.xaxis.set_major_locator(xloc)
## Rotate date labels automatically
fig.autofmt_xdate()
xt.save(train_fig, xt.name(odir, fn+"_prediction" + ("_changepoints" if add_chpts else "") ), c_txt)
### components
train_comp_fig = train.model.plot_components(train.forecast)
ax = plt.gca(); ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
xt.save(train_comp_fig, xt.name(odir, fn+"_components"))
### MAPE
if add_cv:
train_cv_mape = train.cv_mape_fig()
ax = plt.gca()
ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
xt.save(train_cv_mape, xt.name(odir, fn+"_cv_mape"))
plt.tight_layout()
if plot: plt.show()
plt.close('all')
def Save(self):
save(self.storage, self.fname)
def get_new_hessian(self,fname):
path='hessian_collins/hessian_results/'
# get off diagonal parts
F=open(path+fname+'.out')
L=F.readlines()
F.close()
L=[l.strip().split() for l in L]
# construct dict as
# D[par1,par2] = {'P,P':#,'P,M':#,...}
D={}
for l in L:
p1,p2=l[:2]
key='%s,%s'%(p1.replace('M_','').replace('P_',''),p2.replace('M_','').replace('P_',''))
if key not in D: D[key]={}
key2='%s,%s'%(p1.split('_')[0],p2.split('_')[0])
if float(l[-1])>0:
D[key][key2]=float(l[-1])
else:
D[key][key2]=None
# get diagonal parts
F=open(path+fname+'.diag')
L=F.readlines()
F.close()
L=[l.strip().split() for l in L]
for l in L:
p1,p2=l[:2]
key='%s,%s'%(p1.replace('M_','').replace('P_',''),p2.replace('M_','').replace('P_',''))
if key=='FIRST,FIRST':
D['central']=float(l[-1])
continue
if key not in D: D[key]={}
key2='%s,0'%(p1.split('_')[0])
if float(l[-1])>0:
D[key][key2]=float(l[-1])
else:
D[key][key2]=None
# construct hessian
npar=len(self.order)
H=np.zeros((npar,npar))
for k in D:
if k=='central': continue
#print k
#continue
#for kk in D[k]:
# print '\t',kk,D[k][kk]
# get target entry at hess matrix (I,J)
p1,p2=k.split(',')
for i in range(npar):
if self.order[i]==p1: I=i
if self.order[i]==p2: J=i
# step size for target entry
hi=self.par[p1]['err']
hj=self.par[p2]['err']
# compute hess entry
if all([x!=None for x in [D[k][kk] for kk in ['P,P','M,M','P,M','M,P']]]):
#print D[k]['P,P']
#print D[k]['M,M']
#print D[k]['P,M']
#print D[k]['M,P']
#print (4*hi*hj)
if I!=J:
H[I,J]=(D[k]['P,P']+D[k]['M,M']-D[k]['P,M']-D[k]['M,P'])/(4*hi*hj)
elif I==J:
H[I,J]=(D[k]['P,0']+D[k]['M,0']-2*D['central'])/(hi**2)
else:
H[I,J]=None
for i in range(npar):
for j in range(npar):
if i<j: H[j,i]=H[i,j]
# get final hessian = 1/2 dchi2/dxdy
H*=0.5
# print hessian
row='%10s'%''
for i in range(npar): row+='%10s'%self.order[i]
print row
for i in range(npar):
row='%10s '%self.order[i]
for j in range(npar):
if np.isnan(H[i,j])==False:
row+='%10.2e'%H[i,j]
else:
row+='%10s'%''
print row
# save hessian
save(H,'data/'+fname+'.hess')
