In this Saturday morning, I decide to have a look through an ensemble method that has recently driven many successes on Kaggle competitions called “Gradient Boosting Machine”. I then try to reimplement it in python so that I can understand it better in practice.

As its name indicates, GBM trains many models turn by turn and each new model gradually minimises the loss function of the whole system using Gradient Descent method. Assuming each individual modeli is a function h(X;pi) (which we call “base function” or “base learner”) where X is the input and pi is the model parameter. Now let’s choose a loss function L(y,y′) where y is the training output, y′ is the output from the model. In GBM, y′=∑Mi=1βih(X;pi) where M is the number of base learners. What we need to do now is to find:

β,P=argmin{βi,pi}M1L(y,∑i=1Mβih(X;pi))

However this is not easy to achieve optimal parameters. Instead we can try a greedy approach that reduces the loss function stage-by-stage:

βm,pm=argminβ,pL(y,Fm−1(X)+βh(X;p))

And then we update:

Fm=Fm−1(X)+βmh(X;pm)

In order to reduceL(y,Fm) , an obvious way is to step toward the direction where the gradient of L descents:

−gm(X)=–[∂L(y,F(X))∂F(X)]F(X)=Fm−1(X)

However what we want to find out isβm and pm so that Fm∼Fm−1–gm . In another way, βmh(X;pm) is most similar to −gm :

βm,pm=argminβ,p∑i=1N[−gm(xi)–βh(xi;p)]2

We can then fine-tuneβm so that:

βm=argminβ′L(y,Fm−1(X+β′h(x;p)))

I wrote a small python script to demonstrate a simple GBM trainer to learny=xsin(x) over the base function y=xexp(x) using the loss function L(y,y′)=∑i(yi–y′i)2 :

At the end, I would like to recommend Scikit learn to you – a great Python library to work with GBM.

Friedman H J.

As its name indicates, GBM trains many models turn by turn and each new model gradually minimises the loss function of the whole system using Gradient Descent method. Assuming each individual model

However this is not easy to achieve optimal parameters. Instead we can try a greedy approach that reduces the loss function stage-by-stage:

And then we update:

In order to reduce

However what we want to find out is

We can then fine-tune

I wrote a small python script to demonstrate a simple GBM trainer to learn

1
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 |
from numpy import *
from matplotlib import * from scipy.optimize import minimize def gbm(L, dL, p0, M=10): """ Train an ensemble of M base leaners. @param L: callable loss function @param dL: callable loss derivative @param p0: initial guess of parameters @param M: number of base learners """ p = minimize(lambda p: square(-dL(0) - p[0]*h(p[1:])).sum(), p0).x F = p[0]*h(p[1:]) Fs, P, losses = [array(F)], [p], [L(F)] for i in xrange(M): p = minimize(lambda p: square(-dL(F) - p[0]*h(p[1:])).sum(), p0).x p[0] = minimize(lambda a: L(F + a*h(p[1:])), p[0]).x F += p[0]*h(p[1:]) Fs.append(array(F)) P.append(p) losses.append(L(F)) return F, Fs, P, losses X = arange(1, 10, .1) Y = X*sin(X) plot(X, Y) |

1
2 3 4 5 6 7 8 9 10 |
h = lambda a: a[0]*exp(a[1]*X)
L = lambda F: square(Y - F).sum() dL = lambda F: F - Y a0 = asarray([1, 1, 1]) # Build an ensemble of 100 base leaners F, Fs, P, losses = gbm(L, dL, a0, M=100) f = figure(figsize=(10, 10)) _ = plot(X, Y) _ = plot(X, zip(*Fs)) |

1
2 |
#plot losses after each base leaner is added
plot(losses) |

At the end, I would like to recommend Scikit learn to you – a great Python library to work with GBM.

**Bibliography**Friedman H J.

*Greedy Function Approximation: A Gradient Boosting Machine*. IMS 1999 Reitz Lecture. URL:http://www-stat.stanford.edu/~jhf/ftp/trebst.pdf.
## 6 comments :

Hii you are providing good information.Thanks for sharing AND Data Scientist Course in Hyderabad, Data Analytics Courses, Data Science Courses, Business Analytics Training ISB HYD Trained Faculty with 10 yrs of Exp See below link

data-scientist-course-in-hyderabad

Awesome post. After reading your blog I am happy that i got to know new ideas; thanks for sharing this content.

Spoken English Class in Thiruvanmiyur

Spoken English Classes in Adyar

Spoken English Classes in T-Nagar

Spoken English Classes in Vadapalani

Spoken English Classes in Porur

Spoken English Classes in Anna Nagar

Spoken English Classes in Chennai Anna Nagar

Spoken English Classes in Perambur

Spoken English Classes in Anna Nagar West

Great post and informative blog.it was awesome to read, thanks for sharing this great content to my vision.Thanks for sharing your informative post on development.Your work is very good and I appreciate you and hoping for some more informational posts.keep writing and sharing.

Salesforce Training in Chennai

Salesforce Online Training in Chennai

Salesforce Training in Bangalore

Salesforce Training in Hyderabad

Salesforce training in ameerpet

Salesforce Training in Pune

Salesforce Online Training

Salesforce Training

Great job for publishing such a beneficial web site. Your web log isn’t only useful but it is additionally really creative too please do keep sharing more blogs like this.

Cyber Security Training Course in Chennai | Certification | Cyber Security Online Training Course | Ethical Hacking Training Course in Chennai | Certification | Ethical Hacking Online Training Course | CCNA Training Course in Chennai | Certification | CCNA Online Training Course | RPA Robotic Process Automation Training Course in Chennai | Certification | RPA Training Course Chennai | SEO Training in Chennai | Certification | SEO Online Training Course

very good information which is very useful for the readers....thanks for sharing it and do share more posts like this.

WHAT IS NATURAL LANGUAGE PROCESSING? INTRO TO NLP

Happy birthday, my beloved big brother! It seems that the older you grow, the love and care I have for you in my heart grows too.Happy Birthday Wishes For My Brother

Post a Comment