Feels like the gestational period for a software baby. If that’s the case we are what? the size of a grape?
Ten weeks have gone by quickly. Our Team Top-n genera classification network is coming along. I feel as though we have a good design framework and at least know how we are going to approach it. Early on for this project it was apparent that data quantity and quality was going to be a pain point. It looks like we have figured out how we are going to generate labeled training data. The problem of data imbalance is present though, some genera have a 3:1 instance factor so we are going to have to work though manually adding more data or normalizing the set. On the algorithm side we are settled on using the CNN approach first. I have to admit that learning TensorFlow is taking a lot of time and will probably occupy most of the winter break. There are a lot of things happening behind the code that is important to understand, its a very complex system. I have found the book “Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent” by Aurélien Géron very helpful.
The plan for the break is to learn the Tensorflow library better and start developing our V0.0.2 system into a proper ML system. In addition to getting the CNN working I would like to get a transformer network up and running and compare it against a CNN for spectrogram processing. I am only taking the 1 capstone class next term so it will be nice to be able to focus on the project.
Work is crazy busy. We have installed 1 of 2 evaluation tools for our new customer and the product qualification will begin soon. It has been exciting to get to this point but we have a lot of work to do. The next 6 months are really critical and there is a lot of pressure to execute.
Family wise everything is OK. The kids are busy with school and sports as usual. James is doing really well in basketball this winter and Shelby as moved up a lane in swim! Riley is content just doing things she enjoys at home. I need to figure out how to get her to branch out a little more. We also need to either clone ourselves or hire a driver, there always seems to be two or three things going on all at once and no way to get there. Next June Shelby will get her license and that will be a big help or maybe robo-taxi??
We took advantage of the government EV tax incentives before the end of the year and bought a Tesla model Y after thanksgiving. Despite knowing all too well how supervised and un-supervised learning happens, and knowing that self driving autopilot is not actual autopilot, and knowing the number of drivers trapped and killed in weird ways inside EV’s, we still bought one. As an engineer you cant help at marvel at both the hardware and software engineering that goes in to the product. My wife and I both commute 40mi round trip daily for our jobs now and we were buying so much fuel. We will see what the data and savings are after a few months. It may spawn a sub Tesla/EV blog since I love everything automotive.
Blog will resume in…..2025!