def eval(flags):
name = flags.pred_path
yp = pd.read_csv(name)
classes = len([i for i in yp.columns.values if 'class' in i])
yp = yp[['class%d' % i for i in range(1, classes + 1)]].values
myDB = personalDB(flags, name="full")
if "stage1" in name:
y = myDB.data['test_variants_filter']['Class'] - 1
else:
myDB.get_split()
va = myDB.split[flags.fold][1]
y = np.argmax(myDB.y[va], axis=1)
if np.max(y) > classes:
y = np.argmax(to4c(onehot_encode(y)), axis=1)
score = cross_entropy(y, yp)
print(name, score, '\n')
def train_rnn(flags):
myDB = personalDB(flags)
model = BaoRnn(flags, myDB)
model.train()
def train_embedding(flags):
myDB = personalDB(flags)
model = BaoEmbedding(flags, myDB)
model.train()
def show_embedding(flags):
myDB = personalDB(flags)
model = BaoEmbedding(flags, myDB)
model.show_embedding("CBOW/embedding/w:0")
def predict_nn(flags):
myDB = personalDB(flags)
model = BaoNN(flags, myDB)
model.predict()
def build_feature(flags,feas=None):
if feas is None:
feas = ['domain','full_text','gene_text','var_text','share','gene','pattern','onehot']
if 'meta' in flags.task:
feas.append('meta')
myDB = personalDB(flags,name='full')
myDB.get_per_sample_tfidf(['training_text','test_text_filter','stage2_test_text'],"Text")
myDB.get_split()
geneDB = geneTextDB(flags,W=10,bar=0)
geneDB.poke()#_text()
geneDB.get_per_sample_tfidf(['training_text','test_text_filter','stage2_test_text'],"Text")
varDB = geneTextDB(flags,tag='variation',W=10,bar=0)
varDB.poke()#_text()
varDB.get_per_sample_tfidf(['training_text','test_text_filter','stage2_test_text'],"Text")
fold,folds = flags.fold,flags.folds
if fold>=0:
tr_rows,te_rows = myDB.split[fold]
tr_rows = tr_rows.tolist()
te_rows = te_rows.tolist()
else:
tr_rows = None
te_rows = 'stage2' if 'stage2' in flags.task else 'stage1'
words = myDB.select_top_k_words(['training_text','test_text_filter'],"Text",mode="tf",k=5,slack=0)
myDB.get_words(words)
gwords = geneDB.select_top_k_words(['training_text','test_text_filter'],"Text",mode="tf",k=2,slack=0)
geneDB.get_words(gwords)
vwords = varDB.select_top_k_words(['training_text','test_text_filter'],"Text",mode="tf",k=5,slack=0)
varDB.get_words(vwords)
X,Xt = [],[]
if 'meta' in feas:
paths = [#"comps/personal/baobao/backup/91_88_87_86_74",
# "comps/personal/baobao/backup/107_110_105_102_102",
"comps/personal/baobao/backup/4c_52_51_47_46_39"]
for path in paths:
fill(X,Xt,*get_meta(myDB, tr_rows, te_rows, path))
if 'text_len' in feas:
fill(X,Xt,*get_text_len(myDB, tr_rows, te_rows))
if 'domain' in feas:
fill(X,Xt,*domain(myDB, tr_rows, te_rows))
if 'full_text' in feas:
for mode in ["tf"]:
fill(X,Xt,*get_count(myDB, tr_rows, te_rows, mode))
if 'gene_text' in feas:
for mode in ["tf"]:
fill(X,Xt,*get_count(geneDB, tr_rows, te_rows, mode))
if 'var_text' in feas:
for mode in ["tf"]:
fill(X,Xt,*get_count(varDB, tr_rows, te_rows, mode))
if 'share' in feas:
fill(X,Xt,*get_share(myDB, tr_rows, te_rows))
if 'pattern' in feas:
patterns = [r'[a-zA-Z][0-9]+[a-zA-Z]*','del']#,'ins','fus','trunc','methy','amp','sil','expr','splice','exon']
fill(X,Xt,*get_pattern(myDB, tr_rows, te_rows,patterns))
if 'gene' in feas:
fill(X,Xt,*get_gene(myDB, tr_rows, te_rows))
if 'd2v' in feas:
fill(X,Xt,*get_d2v(flags.load_path, tr_rows, te_rows))
if 'onehot' in feas:
fill(X,Xt,*onehot_gene(myDB, tr_rows, te_rows))
y,yt = myDB.gety(tr_rows,te_rows)
return np.hstack(X),y,np.hstack(Xt),yt,te_rows
def preprocess(flags):
myDB = personalDB(flags, name="full")
myDB.poke()
def train_d2v(flags):
myDB = personalDB(flags)
model = D2V(flags, myDB)
model.train()
def show_d2v(flags):
myDB = personalDB(flags)
model = D2V(flags, myDB)
model.show_embedding("D2V/embedding/w:0")
def test_cnn(flags):
myDB = personalDB(flags, name='full')
gDB = geneTextDB(flags, W=10, bar=0)
model = BaoCnn(flags, myDB, gDB)
model.predict()
def train_cnn(flags):
myDB = personalDB(flags, name='full')
gDB = geneTextDB(flags, W=10, bar=0)
model = BaoCnn(flags, myDB, gDB)
model.train()
