The Unofficial Google Data Science Blog

APRIL 23, 2024

To find optimal values of two parameters experimentally, the obvious strategy would be to experiment with and update them in separate, sequential stages. Our experimentation platform supports this kind of grouped-experiments analysis, which allows us to see rough summaries of our designed experiments without much work.

Experimentation 52

Experimentation Optimization Uncertainty Metrics 52

Smart Data Collective

FEBRUARY 4, 2020

Phase 0 is the first to involve human testing. Phase I involves dialing-in the proper dosage and further testing in a larger patient pool. In a report on the failure rates of drug discovery efforts between 2013 and 2015, Richard K. Researching and developing new drugs involves multiple steps called “Phases.”

Experimentation 99

Experimentation Testing Modeling Strategy 99

The Data Visualisation Catalogue

MARCH 18, 2019

Here, 3D Treemaps were tested for visualising the classificatory distribution of patent collections in the International Patent Classification (IPC) system. From exploring all of these examples, you can see that this visualisation isn’t just a wacky experimental way of bringing a Treemap into the third dimension. Luboschnik, H.

Visualization 48

Visualization Interactive Software Experimentation 48

Occam's Razor

APRIL 29, 2013

A benchmark for you: In 2013 if 30% of your time, Ms./Mr. If the simple A/B (test/control) experiment demonstrates that delivering display banner ad impressions to the test group delivers increased revenue, buy impressions to your heart's content. In the latter group I discovered that there were two common themes.

Marketing 166

Marketing Metrics Advertising Measurement 166

Insight

MARCH 12, 2020

In 2013, Robert Galbraith?—?an I tested several different flavors of BERT for use as synopsis classifiers before settling on the DistilBERT model from Hugging Face. On my test set, this approach resulted in~75–95% accuracy and ~.65 Test Case: Dune Let’s see an example of genre tag prediction in action.

Modeling 59

Modeling Metadata Publishing Sales 59

Domino Data Lab

AUGUST 22, 2019

Note: A test set of 19,500 such analogies was developed by Tomas Mikolov and his colleagues in their 2013 word2vec paper. This test set is available at download.tensorflow.org/data/questions-words.txt.]. Relative to extrinsic evaluations, intrinsic tests are quick. Note that the final test word in Table 11.2—ma’am—is

Deep Learning 111

Deep Learning Modeling Metrics Testing 111

Occam's Razor

JANUARY 28, 2013

But there is more to the Security Choice landing page than just passing the all-important transparency test. To accelerate your experimentation I've summarized a cluster of short lessons from experiences covered in this post. So why is it that YouTube's home page is still such a cluttered mess?

Marketing 164

Marketing Insurance Visualization IT 164