Monthly Newsletter — Issue #1
Welcome to my newsletter. I will be summarizing what’s new in Machine Learning, Data Science, and Embedded Systems.
Suggested Reading
I have used macros for so long in the C language, that I never gave it a thought. I am currently writing a book on Database and Communications programming for the Arduino platform, and as this requires quite a bit of code I have been thinking about the quality of code that I will be putting out to the public. This has led me to an article on the need to avoid preprocessor macros.
Next, macro header guards check whether something hasn’t been defined before defining it. Instead, there is the #pragma once instruction that delegates the task of including a header file to the compiler. You can read more about this in this article.
Finally, how do you deal with conflicting variable names in code, especially if those variables are defined in different modules? Namespaces are what you need. If you haven’t been using namespaces, or don’t know why you should, please read this article.
Recent Events
Edge Impulse Imagine conference took place over a period of three days from September 28th. I know that was long ago, but you must check out the session videos. If you don’t know about Edge Impulse, it’s a company/platform that makes it easy to train and deploy ML models on microcontrollers. Some of the sessions that I found to be interesting include: Conservation/AI Ethics Panel; Digital Health and Edge ML; Industrial Automation by Ready Robotics; Computer Vision Techniques for Reducing the Data Collection Pain; Sony’s Spresense Edge Computing with Low-Power Consumption. The third day had some interesting sessions such as: Tiny Robotics, Using MicroROS with Edge Impulse; What’s New with Azure IoT; Machine Learning in the Wild; and, Machine Learning on the Edge with TensorFlow Lite.
Another event that I would recommend catching up on is the Impact Summit organized by Hackster.io. There is a great summary of the sessions available here.
The Google For Africa Event is one that showcases Google’s activities on the continent. It isn’t limited to technical content, but it gives me clarity of purpose in my contributions to the community. For example, I mentor at the Google for Startups Accelerator, and when I see certain startups get highlighted I get a sense of fulfillment. For all of my engagements, I didn’t know about the Hustle Academy until I attended the event.
Google Cloud Next also took place recently. Visit the link to learn all that is new from Google Cloud, as well as what is coming next. On the BigQuery front, it now provides support for new JSON data types, streaming data ingestion with exactly-once execution, integrates with Vertex AI Model Registry, and change data capture with Datastream preview. In preview are: the ability to observe data flow and relevant metrics across stages to troubleshoot query performance, targeted insights on query performance variance compared to historical average, and query monitoring.
BigQuery ML integration with Vertex AI provides capabilities for registering, organizing, versioning, and tracking trained and deployed models. It also makes it possible to store model metadata and runtime dependencies for deployability.
BigQuery has introduced support for unstructured data via something called Object Tables. This lets you create SQL-based pipelines on unstructured data in Google Cloud Storage buckets. Object Tables now make it possible to utilize unstructured data in BQML!
You can jump to this video if you are interested in what’s new for data analysts and data scientists. If you are interested in data engineering then jump here instead.
Embedded Systems
Useful Sensors is a company that has set itself the goal of developing sensors that have Machine Learning baked in. Their first product is called a person sensor and does two things without asking engineers to learn anything new:
- It has a digital pin that goes high when it detects a person in the camera’s field of view
- It has an I2C interface that returns data about the number of persons and their location relative to the sensor
Thanks for reading and keep an eye out for the next issue.
- Robert Thas John
