Introduction
Neural networks are really taking the world by storm. You hear a lot about them, and they are used in your daily life a lot – in fact, the browser that you are reading this very post on probably employs – in part at least – a neural network of some sort. Not only that but neural networks are also used in medical diagnosis, medical image classification, financial predictions, electrical load, energy demand forecasting, process, and quality control. But what do all of these processes have in common? Why are neural networks used for all of these when compared to other algorithms? Well, you can get a sense of that in a video of Cassie Kozyrkov’s talk Making Friends With Machine Learning(link can be found here).
Summary of Video
The talk starts out with Cassie emphasizing how complicated neural networks can sometimes get, thousands of layers with thousands of nodes in each layer. She reminds us that the “proof of the pudding is in the eating”, in other words, the outputs are the things that matter so don’t get bogged down in what actually happens inside, especially if the neural network you are dealing with is really complicated. The talk then gets into the use cases to try neural networks first. They are if simpler methods will fail. For example, really complex tasks like image classification, speech recognition, and language translation. Or if you have quite a bit of domain knowledge in the field and you expect very complicated relationships between the variables.
My Take
Overall, I really liked this video. It was concise and informative. One part I really enjoyed about it was how Cassie reminds us that it is the output that matters, and not to get too bogged down in the weeds of what is actually happening behind the scenes. As I study this subject more it is good to be reminded of that, because I sometimes will try to understand all of the complicated math that is behind this and that really doesn’t matter in most cases as long as you achieve the output. I also really liked the way that Cassie explains all of the concepts, it is easy to understand and follow. One more thing I liked was the visuals that Cassie uses, they are a good mix of fun, and related to the topic. This helps me not only understand the concept that she is teaching but also livens up the talk a little bit to make it more enjoyable.
Conclusion
All in all, I really liked this article. It was informative and concise, while still being fun. I learned a lot and it helped me put some things in context as far as if they matter to the extent to which I thought they mattered. Overall, I highly recommend you watch this video when you get the chance(link to video here).