Turnkey Data Science? Tensorflow Hub is the Big Step

Jason Richards
Nov 5 · 2 min read

As I was going through lessons to earn my Data Science certification, we raw coded every model that has since become a package. From ordinary least square linear regression to deep neural networks, we coded every step in the process to better understand how those models worked.

After attending a walkthrough of Tensorflow 2.0, it occurred to me that methods in the Data Science process are becoming turnkey and Tensorflow Hub is a perfect example.

Tensorlow Hub https://www.tensorflow.org/hub/ is a library of reusable machine learning models. After a simple pip install, you can utilize a vast amount of some of the most popular pre-trained models available. In fact, if you utilize the tensorflow backend, you can have a model like BERT or ELMO run in about four lines of code.

Now, for a person like me, a library this vast with that much computing capability is amazing. Its like being a kid in a candy store. The ability to browse through models for pre-trained text, image and video models is pretty awesome. Then you start to think, “there are programs out there now that can take data, clean it and model it….is my livelihood at risk?”. Simply put, no.

Tensorflow Hub menu

There are two vastly important attributes a Data Scientist has over anyone just throwing data into a machine, that is domain knowledge and experimentation.

Pre-trained models have millions of pieces of information that have been run through GPU and TPU processors to allow a person to apply a smaller dataset to that pre-trained model to get fantastic results. It still, however, takes a scientist to actually research the backstory of the data. Things like industry knowledge, past research in the field and knowing how the data was collected are essential in the Data Science process.

A machine that can take in data and kick out results does have limits. They can only go by what the operator puts in. It takes experimentation, the ability to look at initial results or the steps in the process, reevaluate and retest methods. That is why this particular field is so amazing. Each problem to solve can’t be solved in a turnkey way. Until machines can think like humans, I don’t foresee any turnkey models winning multiple Kaggle competitions or solving problems in different domains.

Jason Richards

Written by

Discovering insights, solving problems and predicting the future through the power of data.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade