Gradient Boosing trains additional models on negative gradient residual

from __future__ import division
import numpy as np
import pandas as pd
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.datasets import load_boston
import ml_metrics as mtr
import matplotlib.pyplot as plt
%matplotlib inline

data = load_boston()
X = data.data
y = data.target

class myGradientBoostingRegression():
    
    def predict(self, X):
        p = np.zeros(X.shape[0])
        for i in range(len(self.learners)):
            dt = self.learners[i]
            c = self.coefs[i]

            p += c*dt.predict(X)
            
        return p
    
    def fit(self, X, y, n_estimators, max_depth):
        self.learners = []
        self.coefs = []
        for n in range(n_estimators):
            if n == 0:
                dt = DecisionTreeRegressor(max_depth=max_depth)
                dt.fit(X,y)
                self.learners.append(dt)
                self.coefs.append(1)
            else:
                p = self.predict(X)
                negtive_gradient_loss = y-p
                
                dt = DecisionTreeRegressor(max_depth=max_depth)
                dt.fit(X, negtive_gradient_loss)
                self.learners.append(dt)
                
                cs = np.arange(0,1,.1)
                losses = [ mtr.mse(p+c*dt.predict(X), y) for c in cs]
                c = cs[np.argmin(losses)]
                self.coefs.append(c)
        
    def score(self, X, y):
        return mtr.mse(self.predict(X), y)
        
        
ns = range(1,10)
ss = []
for n in ns:
    gb = myGradientBoostingRegression()
    gb.fit(X,y,n,3)
    ss.append(gb.score(X,y))
    
plt.plot(ns,ss)