Open RuinChu opened 4 years ago
There seem to be a mistake in calculating risk_score for Cox. In "models/semi_parametric.py" line 308-312:
risk_score = np.exp(np.dot(x, self.weights)) if not use_log: risk_score = np.exp(risk_score) return risk_score
There are a redundant "np.exp()" in line 308 The logged risk score should just be "np.dot(x, self.weights)"
Hope you check it soon. Please tell me if I misunderstand it.
Giving this a bump, seems like the default returns the exp of exp.
There seem to be a mistake in calculating risk_score for Cox. In "models/semi_parametric.py" line 308-312:
There are a redundant "np.exp()" in line 308 The logged risk score should just be "np.dot(x, self.weights)"
Hope you check it soon. Please tell me if I misunderstand it.