When I chose the Fire Risk Prediction as my Capstone project, I knew very little about machine learning. After five weeks of work, I still know very little about the vast field of machine learning, but I know much more about a specific type of neural network, created by our sponsor, Chester Ornes: the visual neural network.
You can read his original publication here. This paper describes the benefits of having a visualization of “the relationship between the query and the data in the training set.” It’s a great way not only to examine the data and the model to understand the relationship, but also to foster a better understanding of how neural networks work.
For me (and the rest of my group), most of these first weeks of the Capstone course have been an intense onboarding process. Our sponsor provided us with some material to start with, but it was up to us to get ourselves up to speed. This has been a pretty painful process for me, if I’m being perfectly honest. As I said earlier, I had basically no knowledge of machine learning concepts beyond that of a layperson.
I spent many, many hours trying to understand this complex topic. Thankfully, there are many resources available for that purpose. The most helpful, unsurprisingly, involved lots and lots of visualizations. One of the first videos I watched did a great job providing dynamic diagrams and graphical representations of concepts they were explaining.
Watching these visual depictions, along with some excellent explanations, really started to drive the knowledge home. I feel like I have a long way to go, but I’m finally able to understand what is happening at a deeper level.
While this isn’t exactly the visualization that we’re working on, reflecting on this experience has definitely reminded me of how valuable visualizations can be in helping people understand complex concepts. It’s exciting to be working on a project that will actually implement this in a way that can help in real world situations, like predicting fire risk.
