tables generation for Tweet experiments
This commit is contained in:
parent
c5ae2f8b1f
commit
43ed808945
|
|
@ -1,11 +1,9 @@
|
|||
import quapy as qp
|
||||
import numpy as np
|
||||
from os import makedirs
|
||||
# from evaluate import evaluate_directory, statistical_significance, get_ranks_from_Gao_Sebastiani
|
||||
import sys, os
|
||||
import pickle
|
||||
from experiments import result_path
|
||||
from result_manager import ResultSet
|
||||
from tabular import Table
|
||||
|
||||
tables_path = './tables'
|
||||
|
|
@ -42,7 +40,6 @@ nice = {
|
|||
}
|
||||
|
||||
|
||||
|
||||
def nicerm(key):
|
||||
return '\mathrm{'+nice[key]+'}'
|
||||
|
||||
|
|
@ -98,13 +95,13 @@ def save_table(path, table):
|
|||
foo.write(table)
|
||||
|
||||
|
||||
|
||||
datasets = qp.datasets.TWITTER_SENTIMENT_DATASETS_TEST
|
||||
evaluation_measures = [qp.error.ae, qp.error.rae]
|
||||
gao_seb_methods = ['cc', 'acc', 'pcc', 'pacc', 'sld', 'svmq', 'svmkld', 'svmnkld']
|
||||
new_methods = []
|
||||
|
||||
def addfunc(dataset, method, loss):
|
||||
|
||||
def experiment_errors(dataset, method, loss):
|
||||
path = result_path(dataset, method, 'm'+loss if not loss.startswith('m') else loss)
|
||||
if os.path.exists(path):
|
||||
true_prevs, estim_prevs, _, _, _, _ = pickle.load(open(path, 'rb'))
|
||||
|
|
@ -127,49 +124,41 @@ for i, eval_func in enumerate(evaluation_measures):
|
|||
nold_methods = len(gao_seb_methods)
|
||||
nnew_methods = len(added_methods)
|
||||
|
||||
table = Table(rows=datasets, cols=methods, addfunc=addfunc)
|
||||
|
||||
# fill table
|
||||
# fill data table
|
||||
table = Table(rows=datasets, cols=methods)
|
||||
for dataset in datasets:
|
||||
for method in methods:
|
||||
table.add(dataset, method, eval_name)
|
||||
table.add(dataset, method, experiment_errors(dataset, method, eval_name))
|
||||
|
||||
# write the latex table
|
||||
tabular = """
|
||||
\\begin{tabularx}{\\textwidth}{|c||""" + ('Y|'*len(gao_seb_methods))+ '|' + ('Y|'*len(added_methods)) + """} \hline
|
||||
& \multicolumn{"""+str(nold_methods)+"""}{c||}{Methods tested in~\cite{Gao:2016uq}} & \multicolumn{"""+str(nnew_methods)+"""}{c|}{} \\\\ \hline
|
||||
\\begin{tabularx}{\\textwidth}{|c||""" + ('Y|'*nold_methods)+ '|' + ('Y|'*nnew_methods) + """} \hline
|
||||
& \multicolumn{"""+str(nold_methods)+"""}{c||}{Methods tested in~\cite{Gao:2016uq}} &
|
||||
\multicolumn{"""+str(nnew_methods)+"""}{c|}{} \\\\ \hline
|
||||
"""
|
||||
|
||||
rowreplace={dataset: nice.get(dataset, dataset.upper()) for dataset in datasets}
|
||||
colreplace={method:'\side{' + nice.get(method, method.upper()) +'$^{' + nicerm(eval_name) + '}$} ' for method in methods}
|
||||
|
||||
tabular += table.latextabular(rowreplace=rowreplace, colreplace=colreplace)
|
||||
tabular += table.latexTabular(rowreplace=rowreplace, colreplace=colreplace)
|
||||
tabular += "\n\end{tabularx}"
|
||||
|
||||
save_table(f'./tables/tab_results_{eval_name}.new2.tex', tabular)
|
||||
|
||||
save_table(f'./tables/tab_results_{eval_name}.new.tex', tabular)
|
||||
|
||||
|
||||
# Tables ranks for AE and RAE (two tables)
|
||||
# ----------------------------------------------------
|
||||
def addfuncRank(dataset, method):
|
||||
rank = table.get(dataset, method, 'rank')
|
||||
if rank is None:
|
||||
return None
|
||||
return [rank]
|
||||
|
||||
methods = gao_seb_methods
|
||||
nold_methods = len(gao_seb_methods)
|
||||
|
||||
ranktable = Table(rows=datasets, cols=methods, addfunc=addfuncRank)
|
||||
# fill table
|
||||
# fill the data table
|
||||
ranktable = Table(rows=datasets, cols=methods, missing='--')
|
||||
for dataset in datasets:
|
||||
for method in methods:
|
||||
ranktable.add(dataset, method)
|
||||
|
||||
ranktable.add(dataset, method, values=table.get(dataset, method, 'rank'))
|
||||
|
||||
# write the latex table
|
||||
tabular = """
|
||||
\\begin{tabularx}{\\textwidth}{|c||""" + ('Y|' * len(gao_seb_methods)) + """} \hline
|
||||
& \multicolumn{""" + str(nold_methods) + """}{c||}{Methods tested in~\cite{Gao:2016uq}} \\\\ \hline
|
||||
& \multicolumn{""" + str(nold_methods) + """}{c|}{Methods tested in~\cite{Gao:2016uq}} \\\\ \hline
|
||||
"""
|
||||
for method in methods:
|
||||
tabular += ' & \side{' + nice.get(method, method.upper()) +'$^{' + nicerm(eval_name) + '}$} '
|
||||
|
|
@ -180,28 +169,29 @@ for i, eval_func in enumerate(evaluation_measures):
|
|||
for method in methods:
|
||||
newrank = ranktable.get(dataset, method)
|
||||
oldrank = gao_seb_ranks[f'{dataset}-{method}-{eval_name}']
|
||||
if newrank is None:
|
||||
newrank = '--'
|
||||
else:
|
||||
newrank = f'{int(newrank)}'
|
||||
tabular += ' & ' + f'{newrank}' + f' ({oldrank}) ' + ranktable.get_color(dataset, method)
|
||||
if newrank != '--':
|
||||
newrank = f'{int(newrank)}'
|
||||
color = ranktable.get_color(dataset, method)
|
||||
if color == '--':
|
||||
color = ''
|
||||
tabular += ' & ' + f'{newrank}' + f' ({oldrank}) ' + color
|
||||
tabular += '\\\\\hline\n'
|
||||
tabular += '\hline\n'
|
||||
|
||||
tabular += 'Average & '
|
||||
tabular += 'Average '
|
||||
for method in methods:
|
||||
newrank = ranktable.get_col_average(method)
|
||||
newrank = ranktable.get_average(method)
|
||||
oldrank = gao_seb_ranks[f'Average-{method}-{eval_name}']
|
||||
if newrank is None or np.isnan(newrank):
|
||||
newrank = '--'
|
||||
else:
|
||||
if newrank != '--':
|
||||
newrank = f'{newrank:.1f}'
|
||||
oldrank = f'{oldrank:.1f}'
|
||||
tabular += ' & ' + f'{newrank}' + f' ({oldrank}) ' + ranktable.get_color(dataset, method)
|
||||
color = ranktable.get_average(method, 'color')
|
||||
if color == '--':
|
||||
color = ''
|
||||
tabular += ' & ' + f'{newrank}' + f' ({oldrank}) ' + color
|
||||
tabular += '\\\\\hline\n'
|
||||
|
||||
tabular += "\end{tabularx}"
|
||||
|
||||
save_table(f'./tables/tab_rank_{eval_name}.new2.tex', tabular)
|
||||
save_table(f'./tables/tab_rank_{eval_name}.new.tex', tabular)
|
||||
|
||||
|
||||
print("[Done]")
|
||||
print("[Done]")
|
||||
|
|
@ -1,204 +0,0 @@
|
|||
from scipy.stats import wilcoxon, ttest_ind_from_stats
|
||||
import numpy as np
|
||||
|
||||
|
||||
|
||||
class ResultSet:
|
||||
VALID_TESTS = [None, "wilcoxon", "ttest_ind_from_stats"]
|
||||
TTEST_DIFF = 'different'
|
||||
TTEST_SIM = 'similar'
|
||||
TTEST_SAME = 'same'
|
||||
|
||||
def __init__(self, name, addfunc, compare='mean', lower_is_better=True, show_std=True, test="wilcoxon",
|
||||
remove_mean='', prec_mean=3, remove_std='', prec_std=3, maxtone=50, minval=None, maxval=None):
|
||||
"""
|
||||
|
||||
:param name: name of the result set (e.g., a Dataset)
|
||||
:param addfunc: a function which is called to process the result input in the "add" method. This function should
|
||||
return a dictionary containing any key-value (e.g., 'mean':0.89) of interest
|
||||
:param compare: the key (as generated by addfunc) that is to be compared in order to rank results
|
||||
:param lower_is_better: if True, lower values of the "compare" key will result in higher ranks
|
||||
:param show_std: whether or not to show the 'std' value (if True, the addfunc is expected to generate it)
|
||||
:param test: which test of statistical significance to use. If "wilcoxon" then scipy.stats.wilcoxon(x,y) will
|
||||
be computed where x,y are the values of the key "values" as computed by addfunc. If "ttest_ind_from_stats", then
|
||||
scipy.stats.ttest_ind_from_stats will be called on "mean", "std", "nobs" values (as computed by addfunc) for
|
||||
both samples being compared.
|
||||
:param remove_mean: if specified, removes the string from the mean (e.g., useful to remove the '0.')
|
||||
:param remove_std: if specified, removes the string from the std (e.g., useful to remove the '0.')
|
||||
"""
|
||||
self.name = name
|
||||
self.addfunc = addfunc
|
||||
self.compare = compare
|
||||
self.lower_is_better = lower_is_better
|
||||
self.show_std = show_std
|
||||
assert test in self.VALID_TESTS, f'unknown test, valid are {self.VALID_TESTS}'
|
||||
self.test = test
|
||||
self.remove_mean = remove_mean
|
||||
self.prec_mean = prec_mean
|
||||
self.remove_std = remove_std
|
||||
self.prec_std = prec_std
|
||||
self.maxtone = maxtone
|
||||
self.minval = minval
|
||||
self.maxval = maxval
|
||||
|
||||
self.r = dict()
|
||||
self.computed = False
|
||||
|
||||
def add(self, key, *args):
|
||||
result = self.addfunc(*args)
|
||||
if result is None:
|
||||
return
|
||||
assert 'values' in result, f'the add function {self.addfunc.__name__} does not fill the "values" attribute'
|
||||
self.r[key] = result
|
||||
vals = self.r[key]['values']
|
||||
if isinstance(vals, np.ndarray):
|
||||
self.r[key]['mean'] = vals.mean()
|
||||
self.r[key]['std'] = vals.std()
|
||||
self.r[key]['nobs'] = len(vals)
|
||||
self.computed = False
|
||||
|
||||
def update(self):
|
||||
if not self.computed:
|
||||
self.compute()
|
||||
|
||||
def compute(self):
|
||||
keylist = np.asarray(list(self.r.keys()))
|
||||
vallist = [self.r[key][self.compare] for key in keylist]
|
||||
keylist = keylist[np.argsort(vallist)]
|
||||
|
||||
print(vallist)
|
||||
self.range_minval = min(vallist) if self.minval is None else self.minval
|
||||
self.range_maxval = max(vallist) if self.maxval is None else self.maxval
|
||||
if not self.lower_is_better:
|
||||
keylist = keylist[::-1]
|
||||
|
||||
# keep track of statistical significance tests; if all are different, then the "phantom dags" will not be shown
|
||||
self.some_similar = False
|
||||
|
||||
for i, key in enumerate(keylist):
|
||||
rank = i + 1
|
||||
isbest = rank == 1
|
||||
if isbest:
|
||||
best = self.r[key]
|
||||
self.r[key]['best'] = isbest
|
||||
self.r[key]['rank'] = rank
|
||||
|
||||
#color
|
||||
val = self.r[key][self.compare]
|
||||
self.r[key]['color'] = self.get_value_color(val, minval=self.range_minval, maxval=self.range_maxval)
|
||||
|
||||
if self.test is not None:
|
||||
if isbest:
|
||||
p_val = 0
|
||||
elif self.test == 'wilcoxon':
|
||||
_, p_val = wilcoxon(best['values'], self.r[key]['values'])
|
||||
elif self.test == 'ttest_ind_from_stats':
|
||||
mean1, std1, nobs1 = best['mean'], best['std'], best['nobs']
|
||||
mean2, std2, nobs2 = self.r[key]['mean'], self.r[key]['std'], self.r[key]['nobs']
|
||||
_, p_val = ttest_ind_from_stats(mean1, std1, nobs1, mean2, std2, nobs2)
|
||||
|
||||
if 0.005 >= p_val:
|
||||
self.r[key]['test'] = ResultSet.TTEST_DIFF
|
||||
elif 0.05 >= p_val > 0.005:
|
||||
self.r[key]['test'] = ResultSet.TTEST_SIM
|
||||
self.some_similar = True
|
||||
elif p_val > 0.05:
|
||||
self.r[key]['test'] = ResultSet.TTEST_SAME
|
||||
self.some_similar = True
|
||||
|
||||
self.computed = True
|
||||
|
||||
def latex(self, key, missing='--', color=True):
|
||||
|
||||
if key not in self.r:
|
||||
return missing
|
||||
|
||||
self.update()
|
||||
|
||||
rd = self.r[key]
|
||||
s = f"{rd['mean']:.{self.prec_mean}f}"
|
||||
if self.remove_mean:
|
||||
s = s.replace(self.remove_mean, '.')
|
||||
if rd['best']:
|
||||
s = "\\textbf{"+s+"}"
|
||||
else:
|
||||
if self.test is not None and self.some_similar:
|
||||
if rd['test'] == ResultSet.TTEST_SIM:
|
||||
s += '^{\dag\phantom{\dag}}'
|
||||
elif rd['test'] == ResultSet.TTEST_SAME:
|
||||
s += '^{\ddag}'
|
||||
elif rd['test'] == ResultSet.TTEST_DIFF:
|
||||
s += '^{\phantom{\ddag}}'
|
||||
|
||||
if self.show_std:
|
||||
std = f"{rd['std']:.{self.prec_std}f}"
|
||||
if self.remove_std:
|
||||
std = std.replace(self.remove_std, '.')
|
||||
s += f" \pm {std}"
|
||||
|
||||
s = f'$ {s} $'
|
||||
if color:
|
||||
s += ' ' + self.r[key]['color']
|
||||
|
||||
return s
|
||||
|
||||
def mean(self, attr='mean', required:int=None, missing=np.nan):
|
||||
"""
|
||||
returns the mean value for the "attr" attribute
|
||||
:param attr: the attribute to average across results
|
||||
:param required: if specified, indicates the number of values that should be part of the mean; if this number
|
||||
is different, then the mean is not computed
|
||||
:param missing: the value to return in case the required condition is not satisfied
|
||||
:return: the mean of the "key" attribute
|
||||
"""
|
||||
keylist = list(self.r.keys())
|
||||
vallist = [self.r[key].get(attr, None) for key in keylist]
|
||||
if None in vallist:
|
||||
return missing
|
||||
if required is not None:
|
||||
if len(vallist) != required:
|
||||
return missing
|
||||
return np.mean(vallist)
|
||||
|
||||
def get(self, key, attr, missing='--'):
|
||||
if key in self.r:
|
||||
self.update()
|
||||
if attr in self.r[key]:
|
||||
return self.r[key][attr]
|
||||
return missing
|
||||
|
||||
def get_color(self, key):
|
||||
if key not in self.r:
|
||||
return ''
|
||||
self.update()
|
||||
return self.r[key]['color']
|
||||
|
||||
def get_value_color(self, val, minval=None, maxval=None):
|
||||
if minval is None or maxval is None:
|
||||
self.update()
|
||||
minval=self.range_minval
|
||||
maxval=self.range_maxval
|
||||
val = (val - minval) / (maxval - minval)
|
||||
if self.lower_is_better:
|
||||
val = 1 - val
|
||||
return color_red2green_01(val, self.maxtone)
|
||||
|
||||
def change_compare(self, attr):
|
||||
self.compare = attr
|
||||
self.computed = False
|
||||
|
||||
|
||||
|
||||
def color_red2green_01(val, maxtone=100):
|
||||
assert 0 <= val <= 1, f'val {val} out of range [0,1]'
|
||||
|
||||
# rescale to [-1,1]
|
||||
val = val * 2 - 1
|
||||
if val < 0:
|
||||
color = 'red'
|
||||
tone = maxtone * (-val)
|
||||
else:
|
||||
color = 'green'
|
||||
tone = maxtone * val
|
||||
return '\cellcolor{' + color + f'!{int(tone)}' + '}'
|
||||
|
||||
|
|
@ -1,247 +0,0 @@
|
|||
import quapy as qp
|
||||
import numpy as np
|
||||
from os import makedirs
|
||||
# from evaluate import evaluate_directory, statistical_significance, get_ranks_from_Gao_Sebastiani
|
||||
import sys, os
|
||||
import pickle
|
||||
from experiments import result_path
|
||||
from result_manager import ResultSet
|
||||
|
||||
|
||||
tables_path = './tables'
|
||||
MAXTONE = 50 # sets the intensity of the maximum color reached by the worst (red) and best (green) results
|
||||
|
||||
makedirs(tables_path, exist_ok=True)
|
||||
|
||||
sample_size = 100
|
||||
qp.environ['SAMPLE_SIZE'] = sample_size
|
||||
|
||||
|
||||
nice = {
|
||||
'mae':'AE',
|
||||
'mrae':'RAE',
|
||||
'ae':'AE',
|
||||
'rae':'RAE',
|
||||
'svmkld': 'SVM(KLD)',
|
||||
'svmnkld': 'SVM(NKLD)',
|
||||
'svmq': 'SVM(Q)',
|
||||
'svmae': 'SVM(AE)',
|
||||
'svmnae': 'SVM(NAE)',
|
||||
'svmmae': 'SVM(AE)',
|
||||
'svmmrae': 'SVM(RAE)',
|
||||
'quanet': 'QuaNet',
|
||||
'hdy': 'HDy',
|
||||
'dys': 'DyS',
|
||||
'svmperf':'',
|
||||
'sanders': 'Sanders',
|
||||
'semeval13': 'SemEval13',
|
||||
'semeval14': 'SemEval14',
|
||||
'semeval15': 'SemEval15',
|
||||
'semeval16': 'SemEval16',
|
||||
'Average': 'Average'
|
||||
}
|
||||
|
||||
|
||||
|
||||
def nicerm(key):
|
||||
return '\mathrm{'+nice[key]+'}'
|
||||
|
||||
def color_from_rel_rank(rel_rank, maxtone=100):
|
||||
rel_rank = rel_rank*2-1
|
||||
if rel_rank < 0:
|
||||
color = 'red'
|
||||
tone = maxtone*(-rel_rank)
|
||||
else:
|
||||
color = 'green'
|
||||
tone = maxtone*rel_rank
|
||||
return '\cellcolor{' + color + f'!{int(tone)}' + '}'
|
||||
|
||||
def color_from_abs_rank(abs_rank, n_methods, maxtone=100):
|
||||
rel_rank = 1.-(abs_rank-1.)/(n_methods-1)
|
||||
return color_from_rel_rank(rel_rank, maxtone)
|
||||
|
||||
|
||||
def load_Gao_Sebastiani_previous_results():
|
||||
def rename(method):
|
||||
old2new = {
|
||||
'kld': 'svmkld',
|
||||
'nkld': 'svmnkld',
|
||||
'qbeta2': 'svmq',
|
||||
'em': 'sld'
|
||||
}
|
||||
return old2new.get(method, method)
|
||||
|
||||
gao_seb_results = {}
|
||||
with open('./Gao_Sebastiani_results.txt', 'rt') as fin:
|
||||
lines = fin.readlines()
|
||||
for line in lines[1:]:
|
||||
line = line.strip()
|
||||
parts = line.lower().split()
|
||||
if len(parts) == 4:
|
||||
dataset, method, ae, rae = parts
|
||||
else:
|
||||
method, ae, rae = parts
|
||||
learner, method = method.split('-')
|
||||
method = rename(method)
|
||||
gao_seb_results[f'{dataset}-{method}-ae'] = float(ae)
|
||||
gao_seb_results[f'{dataset}-{method}-rae'] = float(rae)
|
||||
return gao_seb_results
|
||||
|
||||
|
||||
def get_ranks_from_Gao_Sebastiani():
|
||||
gao_seb_results = load_Gao_Sebastiani_previous_results()
|
||||
datasets = set([key.split('-')[0] for key in gao_seb_results.keys()])
|
||||
methods = np.sort(np.unique([key.split('-')[1] for key in gao_seb_results.keys()]))
|
||||
ranks = {}
|
||||
for metric in ['ae', 'rae']:
|
||||
for dataset in datasets:
|
||||
scores = [gao_seb_results[f'{dataset}-{method}-{metric}'] for method in methods]
|
||||
order = np.argsort(scores)
|
||||
sorted_methods = methods[order]
|
||||
for i, method in enumerate(sorted_methods):
|
||||
ranks[f'{dataset}-{method}-{metric}'] = i+1
|
||||
for method in methods:
|
||||
rankave = np.mean([ranks[f'{dataset}-{method}-{metric}'] for dataset in datasets])
|
||||
ranks[f'Average-{method}-{metric}'] = rankave
|
||||
return ranks, gao_seb_results
|
||||
|
||||
|
||||
def save_table(path, table):
|
||||
print(f'saving results in {path}')
|
||||
with open(path, 'wt') as foo:
|
||||
foo.write(table)
|
||||
|
||||
|
||||
# Tables evaluation scores for AE and RAE (two tables)
|
||||
# ----------------------------------------------------
|
||||
|
||||
datasets = qp.datasets.TWITTER_SENTIMENT_DATASETS_TEST
|
||||
evaluation_measures = [qp.error.ae, qp.error.rae]
|
||||
gao_seb_methods = ['cc', 'acc', 'pcc', 'pacc', 'sld', 'svmq', 'svmkld', 'svmnkld']
|
||||
new_methods = []
|
||||
|
||||
|
||||
def addfunc(dataset, method, loss):
|
||||
path = result_path(dataset, method, 'm'+loss if not loss.startswith('m') else loss)
|
||||
if os.path.exists(path):
|
||||
true_prevs, estim_prevs, _, _, _, _ = pickle.load(open(path, 'rb'))
|
||||
err_fn = getattr(qp.error, loss)
|
||||
errors = err_fn(true_prevs, estim_prevs)
|
||||
return {
|
||||
'values': errors,
|
||||
}
|
||||
return None
|
||||
|
||||
def addave(method, tables):
|
||||
values = []
|
||||
for table in tables:
|
||||
mean = table.get(method, 'values', missing=None)
|
||||
if mean is None:
|
||||
return None
|
||||
values.append(mean)
|
||||
values = np.concatenate(values)
|
||||
return {
|
||||
'values': values
|
||||
}
|
||||
|
||||
def addrankave(method, tables):
|
||||
values = []
|
||||
for table in tables:
|
||||
rank = table.get(method, 'rank', missing=None)
|
||||
if rank is None:
|
||||
return None
|
||||
values.append(rank)
|
||||
return {
|
||||
'values': np.asarray(values)
|
||||
}
|
||||
|
||||
|
||||
TABLES = {eval_func.__name__:{} for eval_func in evaluation_measures}
|
||||
|
||||
for i, eval_func in enumerate(evaluation_measures):
|
||||
eval_name = eval_func.__name__
|
||||
added_methods = ['svm' + eval_name] + new_methods
|
||||
methods = gao_seb_methods + added_methods
|
||||
nold_methods = len(gao_seb_methods)
|
||||
nnew_methods = len(added_methods)
|
||||
|
||||
# fill table
|
||||
TABLE = TABLES[eval_name]
|
||||
for dataset in datasets:
|
||||
TABLE[dataset] = ResultSet(dataset, addfunc, show_std=False, test="ttest_ind_from_stats")
|
||||
for method in methods:
|
||||
TABLE[dataset].add(method, dataset, method, eval_name)
|
||||
|
||||
TABLE['Average'] = ResultSet('ave', addave, show_std=False, test="ttest_ind_from_stats")
|
||||
for method in methods:
|
||||
TABLE['Average'].add(method, method, [TABLE[dataset] for dataset in datasets])
|
||||
|
||||
tabular = """
|
||||
\\begin{tabularx}{\\textwidth}{|c||""" + ('Y|'*len(gao_seb_methods))+ '|' + ('Y|'*len(added_methods)) + """} \hline
|
||||
& \multicolumn{"""+str(nold_methods)+"""}{c||}{Methods tested in~\cite{Gao:2016uq}} & \multicolumn{"""+str(nnew_methods)+"""}{c|}{} \\\\ \hline
|
||||
"""
|
||||
|
||||
for method in methods:
|
||||
tabular += ' & \side{' + nice.get(method, method.upper()) +'$^{' + nicerm(eval_name) + '}$} '
|
||||
tabular += '\\\\\hline\n'
|
||||
|
||||
for dataset in datasets + ['Average']:
|
||||
if dataset == 'Average': tabular+= '\line\n'
|
||||
tabular += nice.get(dataset, dataset.upper()) + ' '
|
||||
for method in methods:
|
||||
tabular += ' & ' + TABLE[dataset].latex(method)
|
||||
tabular += '\\\\\hline\n'
|
||||
|
||||
tabular += "\end{tabularx}"
|
||||
|
||||
save_table(f'./tables/tab_results_{eval_name}.new.tex', tabular)
|
||||
|
||||
|
||||
gao_seb_ranks, gao_seb_results = get_ranks_from_Gao_Sebastiani()
|
||||
|
||||
# Tables ranks for AE and RAE (two tables)
|
||||
# ----------------------------------------------------
|
||||
for i, eval_func in enumerate(evaluation_measures):
|
||||
eval_name = eval_func.__name__
|
||||
methods = gao_seb_methods
|
||||
nold_methods = len(gao_seb_methods)
|
||||
|
||||
TABLE = TABLES[eval_name]
|
||||
TABLE['Average'] = ResultSet('ave', addrankave, show_std=False, test="ttest_ind_from_stats")
|
||||
for method in methods:
|
||||
TABLE['Average'].add(method, method, [TABLE[dataset] for dataset in datasets])
|
||||
|
||||
|
||||
tabular = """
|
||||
\\begin{tabularx}{\\textwidth}{|c||""" + ('Y|' * len(gao_seb_methods)) + """} \hline
|
||||
& \multicolumn{""" + str(nold_methods) + """}{c||}{Methods tested in~\cite{Gao:2016uq}} \\\\ \hline
|
||||
"""
|
||||
|
||||
for method in methods:
|
||||
tabular += ' & \side{' + nice.get(method, method.upper()) +'$^{' + nicerm(eval_name) + '}$} '
|
||||
tabular += '\\\\\hline\n'
|
||||
|
||||
for dataset in datasets + ['Average']:
|
||||
if dataset == 'Average':
|
||||
tabular += '\line\n'
|
||||
else:
|
||||
TABLE[dataset].change_compare('rank')
|
||||
tabular += nice.get(dataset, dataset.upper()) + ' '
|
||||
for method in gao_seb_methods:
|
||||
if dataset == 'Average':
|
||||
method_rank = TABLE[dataset].get(method, 'mean')
|
||||
else:
|
||||
method_rank = TABLE[dataset].get(method, 'rank')
|
||||
gao_seb_rank = gao_seb_ranks[f'{dataset}-{method}-{eval_name}']
|
||||
if dataset == 'Average':
|
||||
if method_rank != '--':
|
||||
method_rank = f'{method_rank:.1f}'
|
||||
gao_seb_rank = f'{gao_seb_rank:.1f}'
|
||||
tabular += ' & ' + f'{method_rank}' + f' ({gao_seb_rank}) ' + TABLE[dataset].get_color(method)
|
||||
tabular += '\\\\\hline\n'
|
||||
tabular += "\end{tabularx}"
|
||||
|
||||
save_table(f'./tables/tab_rank_{eval_name}.new.tex', tabular)
|
||||
|
||||
|
||||
print("[Done]")
|
||||
|
|
@ -6,41 +6,28 @@ from scipy.stats import ttest_ind_from_stats, wilcoxon
|
|||
class Table:
|
||||
VALID_TESTS = [None, "wilcoxon", "ttest"]
|
||||
|
||||
def __init__(self, rows, cols, addfunc, lower_is_better=True, ttest='ttest', prec_mean=3, clean_zero=False,
|
||||
show_std=False, prec_std=3):
|
||||
def __init__(self, rows, cols, lower_is_better=True, ttest='ttest', prec_mean=3,
|
||||
clean_zero=False, show_std=False, prec_std=3, average=True, missing=None, missing_str='--', color=True):
|
||||
assert ttest in self.VALID_TESTS, f'unknown test, valid are {self.VALID_TESTS}'
|
||||
|
||||
self.rows = np.asarray(rows)
|
||||
self.row_index = {row:i for i,row in enumerate(rows)}
|
||||
self.row_index = {row:i for i, row in enumerate(rows)}
|
||||
|
||||
self.cols = np.asarray(cols)
|
||||
self.col_index = {col:j for j,col in enumerate(cols)}
|
||||
self.map = {}
|
||||
self.mfunc = {}
|
||||
self.rarr = {}
|
||||
self.carr = {}
|
||||
self.col_index = {col:j for j, col in enumerate(cols)}
|
||||
|
||||
self.map = {} # keyed (#rows,#cols)-ndarrays holding computations from self.map['values']
|
||||
self._addmap('values', dtype=object)
|
||||
self._addmap('fill', dtype=bool, func=lambda x: x is not None)
|
||||
self._addmap('mean', dtype=float, func=np.mean)
|
||||
self._addmap('std', dtype=float, func=np.std)
|
||||
self._addmap('nobs', dtype=float, func=len)
|
||||
self._addmap('rank', dtype=int, func=None)
|
||||
self._addmap('color', dtype=object, func=None)
|
||||
self._addmap('ttest', dtype=object, func=None)
|
||||
self._addrarr('mean', dtype=float, func=np.mean, argmap='mean')
|
||||
self._addrarr('min', dtype=float, func=np.min, argmap='mean')
|
||||
self._addrarr('max', dtype=float, func=np.max, argmap='mean')
|
||||
self._addcarr('mean', dtype=float, func=np.mean, argmap='mean')
|
||||
self._addcarr('rank-mean', dtype=float, func=np.mean, argmap='rank')
|
||||
if self.nrows>1:
|
||||
self._col_ttest = Table(['ttest'], cols, _merge, lower_is_better, ttest)
|
||||
else:
|
||||
self._col_ttest = None
|
||||
self.addfunc = addfunc
|
||||
self.lower_is_better = lower_is_better
|
||||
self.ttest = ttest
|
||||
self.prec_mean = prec_mean
|
||||
self.clean_zero = clean_zero
|
||||
self.show_std = show_std
|
||||
self.prec_std = prec_std
|
||||
self.add_average = average
|
||||
self.missing = missing
|
||||
self.missing_str = missing_str
|
||||
self.color = color
|
||||
self.touch()
|
||||
|
||||
@property
|
||||
|
|
@ -58,27 +45,6 @@ class Table:
|
|||
if self.modif:
|
||||
self.compute()
|
||||
|
||||
def _addmap(self, map, dtype, func=None):
|
||||
self.map[map] = np.empty((self.nrows, self.ncols), dtype=dtype)
|
||||
self.mfunc[map] = func
|
||||
self.touch()
|
||||
|
||||
def _addrarr(self, rarr, dtype, func=np.mean, argmap='mean'):
|
||||
self.rarr[rarr] = {
|
||||
'arr': np.empty(self.ncols, dtype=dtype),
|
||||
'func': func,
|
||||
'argmap': argmap
|
||||
}
|
||||
self.touch()
|
||||
|
||||
def _addcarr(self, carr, dtype, func=np.mean, argmap='mean'):
|
||||
self.carr[carr] = {
|
||||
'arr': np.empty(self.nrows, dtype=dtype),
|
||||
'func': func,
|
||||
'argmap': argmap
|
||||
}
|
||||
self.touch()
|
||||
|
||||
def _getfilled(self):
|
||||
return np.argwhere(self.map['fill'])
|
||||
|
||||
|
|
@ -89,34 +55,19 @@ class Table:
|
|||
def _indexes(self):
|
||||
return itertools.product(range(self.nrows), range(self.ncols))
|
||||
|
||||
def _runmap(self, map):
|
||||
def _addmap(self, map, dtype, func=None):
|
||||
self.map[map] = np.empty((self.nrows, self.ncols), dtype=dtype)
|
||||
if func is None:
|
||||
return
|
||||
m = self.map[map]
|
||||
f = self.mfunc[map]
|
||||
f = func
|
||||
if f is None:
|
||||
return
|
||||
indexes = self._indexes() if map == 'fill' else self._getfilled()
|
||||
for i,j in indexes:
|
||||
m[i,j] = f(self.values[i,j])
|
||||
for i, j in indexes:
|
||||
m[i, j] = f(self.values[i, j])
|
||||
|
||||
def _runrarr(self, rarr):
|
||||
dic = self.rarr[rarr]
|
||||
arr, f, map = dic['arr'], dic['func'], dic['argmap']
|
||||
for col, cid in self.col_index.items():
|
||||
if all(self.map['fill'][:, cid]):
|
||||
arr[cid] = f(self.map[map][:, cid])
|
||||
else:
|
||||
arr[cid] = None
|
||||
|
||||
def _runcarr(self, carr):
|
||||
dic = self.carr[carr]
|
||||
arr, f, map = dic['arr'], dic['func'], dic['argmap']
|
||||
for row, rid in self.row_index.items():
|
||||
if all(self.map['fill'][rid, :]):
|
||||
arr[rid] = f(self.map[map][rid, :])
|
||||
else:
|
||||
arr[rid] = None
|
||||
|
||||
def _runrank(self):
|
||||
def _addrank(self):
|
||||
for i in range(self.nrows):
|
||||
filled_cols_idx = np.argwhere(self.map['fill'][i]).flatten()
|
||||
col_means = [self.map['mean'][i,j] for j in filled_cols_idx]
|
||||
|
|
@ -125,7 +76,7 @@ class Table:
|
|||
ranked_cols_idx = ranked_cols_idx[::-1]
|
||||
self.map['rank'][i, ranked_cols_idx] = np.arange(1, len(filled_cols_idx)+1)
|
||||
|
||||
def _runcolor(self):
|
||||
def _addcolor(self):
|
||||
for i in range(self.nrows):
|
||||
filled_cols_idx = np.argwhere(self.map['fill'][i]).flatten()
|
||||
if filled_cols_idx.size==0:
|
||||
|
|
@ -144,6 +95,12 @@ class Table:
|
|||
normval = 1 - normval
|
||||
self.map['color'][i, col_idx] = color_red2green_01(normval)
|
||||
|
||||
def _addlatex(self):
|
||||
return
|
||||
for i,j in self._getfilled():
|
||||
self.map['latex'][i,j] = self.latex(self.rows[i], self.cols[j])
|
||||
|
||||
|
||||
def _run_ttest(self, row, col1, col2):
|
||||
mean1 = self.map['mean'][row, col1]
|
||||
std1 = self.map['std'][row, col1]
|
||||
|
|
@ -160,10 +117,10 @@ class Table:
|
|||
_, p_val = wilcoxon(values1, values2)
|
||||
return p_val
|
||||
|
||||
def _runttest(self):
|
||||
def _addttest(self):
|
||||
if self.ttest is None:
|
||||
return
|
||||
self.some_similar = False
|
||||
self.some_similar = [False]*self.ncols
|
||||
for i in range(self.nrows):
|
||||
filled_cols_idx = np.argwhere(self.map['fill'][i]).flatten()
|
||||
if len(filled_cols_idx) <= 1:
|
||||
|
|
@ -182,62 +139,74 @@ class Table:
|
|||
pval_outcome = pval_interpretation(p_val)
|
||||
self.map['ttest'][i, j] = pval_outcome
|
||||
if pval_outcome != 'Diff':
|
||||
self.some_similar = True
|
||||
|
||||
def get_col_average(self, col, arr='mean'):
|
||||
self.update()
|
||||
cid = self.col_index[col]
|
||||
return self.rarr[arr]['arr'][cid]
|
||||
|
||||
def _map_list(self):
|
||||
maps = list(self.map.keys())
|
||||
maps.remove('fill')
|
||||
maps.remove('values')
|
||||
maps.remove('color')
|
||||
maps.remove('ttest')
|
||||
return ['fill'] + maps
|
||||
self.some_similar[j] = True
|
||||
|
||||
def compute(self):
|
||||
for map in self._map_list():
|
||||
self._runmap(map)
|
||||
self._runrank()
|
||||
self._runcolor()
|
||||
self._runttest()
|
||||
for arr in self.rarr.keys():
|
||||
self._runrarr(arr)
|
||||
for arr in self.carr.keys():
|
||||
self._runcarr(arr)
|
||||
if self._col_ttest != None:
|
||||
for col in self.cols:
|
||||
self._col_ttest.add('ttest', col, self.col_index[col], self.map['fill'], self.values, self.map['mean'], self.ttest)
|
||||
self._col_ttest.compute()
|
||||
self._addmap('fill', dtype=bool, func=lambda x: x is not None)
|
||||
self._addmap('mean', dtype=float, func=np.mean)
|
||||
self._addmap('std', dtype=float, func=np.std)
|
||||
self._addmap('nobs', dtype=float, func=len)
|
||||
self._addmap('rank', dtype=int, func=None)
|
||||
self._addmap('color', dtype=object, func=None)
|
||||
self._addmap('ttest', dtype=object, func=None)
|
||||
self._addmap('latex', dtype=object, func=None)
|
||||
self._addrank()
|
||||
self._addcolor()
|
||||
self._addttest()
|
||||
self._addlatex()
|
||||
if self.add_average:
|
||||
self._addave()
|
||||
self.modif = False
|
||||
|
||||
def add(self, row, col, *args, **kwargs):
|
||||
print(row, col, args, kwargs)
|
||||
values = self.addfunc(row, col, *args, **kwargs)
|
||||
# if values is None:
|
||||
# raise ValueError(f'addfunc returned None for row={row} col={col}')
|
||||
rid, cid = self.coord(row, col)
|
||||
def _is_column_full(self, col):
|
||||
return all(self.map['fill'][:, self.col_index[col]])
|
||||
|
||||
def _addave(self):
|
||||
ave = Table(['ave'], self.cols, lower_is_better=self.lower_is_better, ttest=self.ttest, average=False,
|
||||
missing=self.missing, missing_str=self.missing_str)
|
||||
for col in self.cols:
|
||||
values = None
|
||||
if self._is_column_full(col):
|
||||
if self.ttest == 'ttest':
|
||||
values = np.asarray(self.map['mean'][:, self.col_index[col]])
|
||||
else: # wilcoxon
|
||||
values = np.concatenate(self.values[:, self.col_index[col]])
|
||||
ave.add('ave', col, values)
|
||||
self.average = ave
|
||||
|
||||
def add(self, row, col, values):
|
||||
if values is not None:
|
||||
values = np.asarray(values)
|
||||
if values.ndim==0:
|
||||
values = values.flatten()
|
||||
rid, cid = self._coordinates(row, col)
|
||||
self.map['values'][rid, cid] = values
|
||||
self.touch()
|
||||
|
||||
def get(self, row, col, attr='mean'):
|
||||
assert attr in self.map, f'unknwon attribute {attr}'
|
||||
self.update()
|
||||
rid, cid = self.coord(row, col)
|
||||
assert attr in self.map, f'unknwon attribute {attr}'
|
||||
rid, cid = self._coordinates(row, col)
|
||||
if self.map['fill'][rid, cid]:
|
||||
return self.map[attr][rid, cid]
|
||||
v = self.map[attr][rid, cid]
|
||||
if v is None or (isinstance(v,float) and np.isnan(v)):
|
||||
return self.missing
|
||||
return v
|
||||
else:
|
||||
return self.missing
|
||||
|
||||
def coord(self, row, col):
|
||||
def _coordinates(self, row, col):
|
||||
assert row in self.row_index, f'row {row} out of range'
|
||||
assert col in self.col_index, f'col {col} out of range'
|
||||
rid = self.row_index[row]
|
||||
cid = self.col_index[col]
|
||||
return rid, cid
|
||||
|
||||
def get_col_table(self):
|
||||
return self._col_ttest
|
||||
def get_average(self, col, attr='mean'):
|
||||
self.update()
|
||||
if self.add_average:
|
||||
return self.average.get('ave', col, attr=attr)
|
||||
return None
|
||||
|
||||
def get_color(self, row, col):
|
||||
color = self.get(row, col, attr='color')
|
||||
|
|
@ -245,11 +214,11 @@ class Table:
|
|||
return ''
|
||||
return color
|
||||
|
||||
def latex(self, row, col, missing='--', color=True):
|
||||
def latex(self, row, col):
|
||||
self.update()
|
||||
i,j = self.coord(row, col)
|
||||
i,j = self._coordinates(row, col)
|
||||
if self.map['fill'][i,j] == False:
|
||||
return missing
|
||||
return self.missing_str
|
||||
|
||||
mean = self.map['mean'][i,j]
|
||||
l = f" {mean:.{self.prec_mean}f}"
|
||||
|
|
@ -257,78 +226,69 @@ class Table:
|
|||
l = l.replace(' 0.', '.')
|
||||
|
||||
isbest = self.map['rank'][i,j] == 1
|
||||
|
||||
if isbest:
|
||||
l = "\\textbf{"+l+"}"
|
||||
else:
|
||||
if self.ttest is not None and self.some_similar:
|
||||
test_label = self.map['ttest'][i,j]
|
||||
if test_label == 'Sim':
|
||||
l += '^{\dag\phantom{\dag}}'
|
||||
elif test_label == 'Same':
|
||||
l += '^{\ddag}'
|
||||
elif test_label == 'Diff':
|
||||
l += '^{\phantom{\ddag}}'
|
||||
l = "\\textbf{"+l.strip()+"}"
|
||||
|
||||
stat = ''
|
||||
if self.ttest is not None and self.some_similar[j]:
|
||||
test_label = self.map['ttest'][i,j]
|
||||
if test_label == 'Sim':
|
||||
stat = '^{\dag\phantom{\dag}}'
|
||||
elif test_label == 'Same':
|
||||
stat = '^{\ddag}'
|
||||
elif isbest or test_label == 'Diff':
|
||||
stat = '^{\phantom{\ddag}}'
|
||||
|
||||
std = ''
|
||||
if self.show_std:
|
||||
std = self.map['std'][i,j]
|
||||
std = f" {std:.{self.prec_std}f}"
|
||||
if self.clean_zero:
|
||||
std = std.replace(' 0.', '.')
|
||||
l += f" \pm {std}"
|
||||
std = f" \pm {std:{self.prec_std}}"
|
||||
|
||||
l = f'$ {l} $'
|
||||
if color:
|
||||
if stat!='' or std!='':
|
||||
l = f'{l}${stat}{std}$'
|
||||
|
||||
if self.color:
|
||||
l += ' ' + self.map['color'][i,j]
|
||||
|
||||
return l
|
||||
|
||||
def latextabular(self, missing='--', color=True, rowreplace={}, colreplace={}, average=True):
|
||||
def latexTabular(self, rowreplace={}, colreplace={}, average=True):
|
||||
tab = ' & '
|
||||
tab += ' & '.join([colreplace.get(col, col) for col in self.cols])
|
||||
tab += ' \\\\\hline\n'
|
||||
for row in self.rows:
|
||||
rowname = rowreplace.get(row, row)
|
||||
tab += rowname + ' & '
|
||||
tab += self.latexrow(row, missing, color)
|
||||
tab += ' \\\\\hline\n'
|
||||
tab += self.latexRow(row)
|
||||
|
||||
if average:
|
||||
tab += '\hline\n'
|
||||
tab += 'Average & '
|
||||
tab += self.latexave(missing, color)
|
||||
tab += ' \\\\\hline\n'
|
||||
tab += self.latexAverage()
|
||||
return tab
|
||||
|
||||
|
||||
def latexrow(self, row, missing='--', color=True):
|
||||
s = [self.latex(row, col, missing=missing, color=color) for col in self.cols]
|
||||
def latexRow(self, row, endl='\\\\\hline\n'):
|
||||
s = [self.latex(row, col) for col in self.cols]
|
||||
s = ' & '.join(s)
|
||||
s += ' ' + endl
|
||||
return s
|
||||
|
||||
def latexave(self, missing='--', color=True):
|
||||
return self._col_ttest.latexrow('ttest')
|
||||
def latexAverage(self, endl='\\\\\hline\n'):
|
||||
if self.add_average:
|
||||
return self.average.latexRow('ave', endl=endl)
|
||||
|
||||
def get_rank_table(self):
|
||||
t = Table(rows=self.rows, cols=self.cols, addfunc=_getrank, ttest=None, prec_mean=0)
|
||||
for row, col in self._getfilled():
|
||||
t.add(self.rows[row], self.cols[col], row, col, self.map['rank'])
|
||||
def getRankTable(self):
|
||||
t = Table(rows=self.rows, cols=self.cols, prec_mean=0, average=True)
|
||||
for rid, cid in self._getfilled():
|
||||
row = self.rows[rid]
|
||||
col = self.cols[cid]
|
||||
t.add(row, col, self.get(row, col, 'rank'))
|
||||
t.compute()
|
||||
return t
|
||||
|
||||
def _getrank(row, col, rowid, colid, rank):
|
||||
return [rank[rowid, colid]]
|
||||
|
||||
def _merge(unused, col, colidx, fill, values, means, ttest):
|
||||
if all(fill[:,colidx]):
|
||||
nrows = values.shape[0]
|
||||
if ttest=='ttest':
|
||||
values = np.asarray(means[:, colidx])
|
||||
else: # wilcoxon
|
||||
values = [values[i, colidx] for i in range(nrows)]
|
||||
values = np.concatenate(values)
|
||||
return values
|
||||
else:
|
||||
return None
|
||||
|
||||
def pval_interpretation(p_val):
|
||||
if 0.005 >= p_val:
|
||||
return 'Diff'
|
||||
|
|
@ -352,21 +312,3 @@ def color_red2green_01(val, maxtone=50):
|
|||
tone = maxtone * val
|
||||
return '\cellcolor{' + color + f'!{int(tone)}' + '}'
|
||||
|
||||
#
|
||||
# def addfunc(m,d, mean, size):
|
||||
# return np.random.rand(size)+mean
|
||||
#
|
||||
# t = Table(rows = ['M1', 'M2', 'M3'], cols=['D1', 'D2', 'D3', 'D4'], addfunc=addfunc, ttest='wilcoxon')
|
||||
# t.add('M1','D1', mean=0.5, size=100)
|
||||
# t.add('M1','D2', mean=0.5, size=100)
|
||||
# t.add('M2','D1', mean=0.2, size=100)
|
||||
# t.add('M2','D2', mean=0.1, size=100)
|
||||
# t.add('M2','D3', mean=0.7, size=100)
|
||||
# t.add('M2','D4', mean=0.3, size=100)
|
||||
# t.add('M3','D1', mean=0.9, size=100)
|
||||
# t.add('M3','D2', mean=0, size=100)
|
||||
#
|
||||
# print(t.latextabular())
|
||||
#
|
||||
# print('rank')
|
||||
# print(t.get_rank_table().latextabular())
|
||||
|
|
|
|||
Loading…
Reference in New Issue