Press "Enter" to skip to content

Day: January 12, 2021

Hyperparameter Tuning as Technical Debt

John Mount has an interesting take on hyperparameter tuning:

The hyper dance is the venial trick of pushing user facing technical debt and flaws as user controllable features. These controls are usually named “hyper parameters” and they are parameters or arguments that control the behavior of an algorithm. Users think “hyper parameters” must be even better than “regular parameters”, just like “hyper drive” is better than “sub-light drive.” However the etymology of the name isn’t from science fiction, it is just the need in statistical contexts to have a name for controls other than parameter, as parameter is often used to name the fit coefficients of a model (i.e. to name an output, not an input!).

In addition to this, I’d be concerned that heavy hyperparameter tuning could lead to a garden of forking paths problem where we end up accidentally doing the equivalent of p-hacking: modifying hyperparameters until we come up with the “right” answer.

Comments closed

Improving a Graph

Elizabeth Ricks has started a series on improving a particular visual:

I empathize with the plight of this anonymous creator. In previous roles, I frequently created visuals that looked like this, and was left frustrated when requests came back for “more data.” I slowly came to realize that I was assigning my audience the tedious task of figuring out for themselves what the takeaways were. My visuals should have been highlighting the interesting things to those seeing them for the first time. The five questions we’ll be discussing in this series will help us to do just that.

The first question in the series is, “What elements can I eliminate?” I think that’s a really good idea—with data visualization, less is more.

Comments closed

Archival on Delete in SQL Server

Erik Darling shows off a pattern:

Well, friends, I have good news for you. This is an easy one to implement.

Let’s say that in Stack Overflow land, when a user deletes their account we also delete all their votes. That’s not how it works, but it’s how I’m going to show you how to condense what can normally be a difficult process to isolate into a single operation.

The one gripe I have with this post is that my annoyingly loud keyboard is buckling spring, not Cherry MX Blue, thank-you-very-much.

Comments closed

Azure Data Factory and Source Control

Ahmad Yaseen shows how you can save Azure Data Factory pipelines in source control:

To overcome these limitations, Azure Data Factory provides us with the ability to integrate with a GIT repository, such as Azure DevOps or GitHub repository, that helps in tracking and versioning the pipelines changes, and incrementally save the pipeline changes during the development stage, without the need to validate the incomplete pipeline, preventing these changes from being lost in case of any crash or failure. In this case, you will be able to test the pipeline, revert any change that is detected as a bug, and publish the pipeline to the Data Factory when everything is developed and validated successfully.

Click through for the setup instructions.

Comments closed

Query Store and Cross-Database Queries

Matthew McGiffen does some research:

When I was writing the script shared in my last post Identify the (Top 20) most expensive queries across your SQL Server using Query Store a question crossed my mind:

Query Store is a configuration that is enabled per database, and the plans and stats for queries executed in that database are stored in the database itself. So what does query store do when a query spans more than one database?

Read on for the answer.

Comments closed