10_print_perf_layer.py

import pickle
import pandas as pd
import numpy as np
from sklearn.metrics import accuracy_score, f1_score, roc_auc_score


layers=[1,2,4,8,16]

for layer in layers:
    
    results = pickle.load(open(f"final_MLP/emb_8/layer_{layer}.pickle",'rb'))
    
    events = list(results[0].keys())
    
    accs = []
    f1s = []
    aucs = []
    
    for seed, data in results.items():  
        
        y_true = []
        y_pred = []
        y_score = []
        
        for testEv in events:
            
            y_true += data[testEv]["y_true"]
            y_pred += data[testEv]["y_pred"]
            y_score += data[testEv]["y_score"]
            
        
        accs.append(accuracy_score(y_true, y_pred))
        f1s.append(f1_score(y_true, y_pred))
        aucs.append(roc_auc_score(y_true, y_score))     
        
    metrics = {
        "Accuracy (Mean)": np.array(accs).mean(),
        "Accuracy (Std dev)": np.array(accs).std(),
        "F1 score (Mean)": np.array(f1s).mean(),
        "F1 score (Std dev)": np.array(f1s).std(),
        "Auc (Mean)": np.array(aucs).mean(),
        "Auc (Std dev)": np.array(aucs).std(),
    }
    
    print("\\textbf{"+str(layer)+"} & \\textbf{%.2f} & \\textbf{%.2f} & \\textbf{%.2f} & \\textbf{%.2f} & \\textbf{%.2f} & \\textbf{%.2f}  \\\\ \hline" % (metrics["Accuracy (Mean)"], metrics["Accuracy (Std dev)"],metrics["F1 score (Mean)"],metrics["F1 score (Std dev)"], metrics["Auc (Mean)"],metrics["Auc (Std dev)"]))