datasets = ens.read_dataset(paths["common"], paths["binary"])
helper = get_renam_fun(json_path)
if (get_fun):
return helper
new_datasets = datasets #[helper(data_i) for data_i in datasets]
votes = ens.Votes(learn.train_ens(new_datasets, clf="LR"))
result_i = votes.voting(False)
result_i.report()
def get_renam_fun(json_path):
rename = read_rename(json_path)
def helper(data_i):
feat_i = feats.Feats()
for name_i, rename_i in rename.items():
print((rename_i, name_i))
feat_i[rename_i] = data_i[name_i]
return feat_i
return helper
if __name__ == "__main__":
dataset = "3DHOI"
dir_path = ".."
paths = exp.basic_paths(dataset, dir_path, "dtw", None)
paths["common"] = ["../3DHOI/1D_CNN/feats"]
rename_frames(paths, "rename")
# rename=random_cat("../3DHOI/1D_CNN/feats")
# save_rename("rename",rename)
exp.result_exp(loss_name,result_dict,out_i)
def weight_desc(result_dict,eps=0.02):
weight_dict={}
for name_i,pair_i in result_dict.items():
result_i,weights_i=pair_i
s_clf=(weights_i>eps)
n_clf=weights_i[s_clf].shape[0]
s_clf=str(np.where(s_clf==True)[0])
new_name_i="%s,%d,%s" % (name_i,n_clf,s_clf)
weight_dict[new_name_i]=result_i
return weight_dict
def single_exp(paths,loss_type,out_path,p=0.5,k=10):
loss_dict={"MSE":MSE,"gasen":gasen.Corl,"Comb":Comb}
valid=auc.CrossVal(p)
if(p<1.0):
valid=auc.MedianaVal(valid,k=k)
optim=OptimWeights(loss_dict[loss_type],valid)
result=optim(paths)[0]
result.report()
result.get_cf(out_path)
if __name__ == "__main__":
dataset="MHAD"
dir_path="../../ICSS"#%s" % dataset
paths=exp.basic_paths(dataset,dir_path,"dtw","ens/feats")
paths["common"].append("%s/%s/1D_CNN/feats" % (dir_path,dataset))
print(paths)
# optim=auc_exp(paths,"MHAD")
single_exp(paths,"Comb","cf/%s" % dataset,p=1.0)
