Press "Enter" to skip to content

Day: September 8, 2025

An Introduction to Batch Normalization in Neural Networks

Published 2025-09-08 by Kevin Feasel

Ivan Palomares Carrascosa shows off one technique for optimizing neural networks:

Deep neural networks have drastically evolved over the years, overcoming common challenges that arise when training these complex models. This evolution has enabled them to solve increasingly difficult problems effectively.

One of the mechanisms that has proven especially influential in the advancement of neural network-based models is batch normalization. This article provides a gentle introduction to this strategy, which has become a standard in many modern architectures, helping to improve model performance by stabilizing training, speeding up convergence, and more.

Read on for a quick description of how it works and a demonstration in Keras.

Leave a Comment

Enabling Map Visuals in Power BI

Published 2025-09-08 by Kevin Feasel

Boniface Muchendu gets past the X:

Have you ever tried to create a map in Power BI only to see an error instead of your visualization? If your Power BI maps are not working, you’re not alone. By default, some map and filled map visuals may be disabled due to security settings. The good news? With a few quick adjustments, you can enable maps in Power BI Desktop or, if needed, in your organization’s tenant settings.

Read on to see why this visual might be disabled and how to enable it.

Leave a Comment

The Consequences of Hitting Semantic Model Guardrails

Published 2025-09-08 by Kevin Feasel

Chris Webb smashes into a wall:

Direct Lake mode in Power BI allows you to build semantic models on very large volumes of data, but because it is still an in-memory database engine there are limits on how much data it can work with. As a result it has rules – called guardrails – that it uses to check whether you are trying to build a semantic model that is too large. But what happens when you hit those guardrails? This week one of my colleagues, Gaurav Agarwal, showed me the results of some tests that he did which I thought I would share here.

Click through to see what happens when you go past one of those guardrails.

Leave a Comment

Load Testing SQL Server with HammerDB and Docker

Published 2025-09-08 by Kevin Feasel

Anthony Nocentino announces a new tool:

I’m excited to announce the release of a new open-source project that fully automates HammerDB benchmarking for SQL Server using Docker. If you’ve ever needed to run TPC-C or TPC-H benchmarks multiple times, you know how time-consuming the manual setup can be. This project removes the hassle and gets you up and running a single command: ./loadtest.sh.

Click through to learn more about the project and how you can grab the code.

Leave a Comment

Batching Large Data Operations via Key Ranges

Published 2025-09-08 by Kevin Feasel

Andy Brownsword updates or deletes a batch of rows:

Effective batching in general helps us by:

  • Reduce transaction length and minimise blocking
  • Avoids unnecessary checking of the same rows repeatedly
  • Introduce graceful pacing to reduce impact on busy environments or data replication

I’m not the biggest fan of the OFFSET/FETCH combination there, at least if your key column is fairly well packed—like, say, 99+% of the rows are contiguous and you occasionally have a jump of a few thousand rows. Also, that batch size of 100K might be a little high, although that will certainly depend on what the operation is. Batch updating a column based on some fairly straightforward calculation? You can probably get away with 100K, though I’d still prefer 10K. But as you add more complexities (deleting rows, very high server throughput, triggers, limited hardware, etc.), that number should edge downward.

Leave a Comment