1. Introduction

Let us start the story with a data science project that predicts users credit scoring using the telco data at country A. It is a successful machine learning (ML) project as we have sufficient large and comprehensive labelled data.

Then, the business team expects the data science team to duplicate the same model for a new market at country B quickly. However, we can’t as the currency and consumers behaviour of country B are different from country A.

I believe the story is common in data science companies. Generally, the ML model is built on the assumption that the training and test data are extracted from same feature space and same distribution. In other words, once the distribution shifts, the model fails.

Once the distribution shifts, the model fails.

Researchers have a long thought of this problem with the solution called transfer learning. In a layman term, we have labelled data from the source domain and we would like to build a ML model for the target domain of different tasks or distribution than the source domain (Pan and Yang, 2010).

In this article, we will experiment on a transfer learning method that proposed by Hal Daume III (2006), named easy adaptation (this name is coined in his later paper). In the followings, we will briefly explain easy adaptation in Section 2 and the experiment in Section 3. Finally, the conclusion is drawn in Section 4.

2. Transfer Learning with Easy Adaptation

Easy adaptation has a simple construction method in Daumé III paper of title “Frustratingly Easy Domain Adaptation”. Say that we have labelled data of similar feature space (attributes of x0 and x1 with output y) in both source and target domains but in different distribution (refer to Figure 1). For instance, the second record from source domain data is (x0=2, x1=20, y=2) but the output becomes y=1 in the first record of target domain data.

Continue reading “Transfer Learning – Makes the Machine Learning Models Works Even with Insufficient* Labelled Data” →

[My original post from linkedin https://www.linkedin.com/pulse/things-job-title-zan-kai-chong/%5D

Switching from the job title data scientist to machine learning engineer amuses a lot of my friends. They wondering since when I become a lecturer again (note : machine “learning”). Despite my wrongful explanation, I start thinking what is my real job title other than those words printed on my name cards.

Analytically, I should list down all my job functions. Then build the heat map or histogram from all the words in the description and then identify the common words by applying the max(count) or corr function. Okay, sounds right. Here we go.

First, I work on AWS platform. As a trustworthy power user, I stress-test the costly computing instance and provide my helpful IT support to new comers (lady is preferable). I also use EMR (very expensive computer clusters) like-a-boss occasionally for big data stuff. Occasionally, I speak AWS jargon as if I am real AWS engineer.

Okay. You saw the word “big data”. Of cause , I am (acting like) a big data engineer as I work on peta-ful (new word to describe peta bytes) data. These petaful data are our asset to track you. We may not be as good as Cambridge Analytica. But we know many things about you and what you did last weekend. The more you attach to your phone, the more we know you.

In short , it is fun to work in analytics company. Well, my real job title? I am wondering as well . Perhaps , I should just call myself “engineer“.

I was in a great self-involved until my civil engineer friends start mocking me with a photo.