A Warm Welcome To ML/AI for Gem City Tech
Innovations in machine learning, artificial intelligence, and data analysis are pushing the tech world forward. That is why it is exciting that Gem City Tech has started a subgroup under their umbrella focused exclusively on Machine Learning and Artificial Intelligence, with additional room for exploration of topics in data presentation and analytics.
Tonight was the first meeting of this subgroup, which means the timing of finding Gem City Tech could not have been any better, given that this is a topic that I would like to explore in my career. Like my previous meet-up with Gem City Tech, I felt a bit overwhelmed, but in a good way, it’s like trying to drink out of a firehose. The information is coming at me quickly, but it remains informative and understandable.
When it comes to the vast ocean of Machine Learning, I find myself in the metaphorical kiddie pool. My feet are wet enough to understand some of the general topics, but there is so much that I need to learn, and one of the essential parts of learning is feedback. The opportunity to connect with like-minded individuals on this topic will give me the correct feedback to guide me in the right direction.
Feedback is a crucial step not only for the learning journey but is an essential step for this group’s journey as well. This first meeting was just a feeling-out process; everyone in attendance gave feedback about where the group stood regarding ML knowledge. I was not alone in my feeling of information overload. And it was not because of our speaker, Evelyn Boettcher. They did a fantastic job and were excited about how ML worked.
As a group, we did an example project to answer the question of whether randomness is genuinely random?
The first example that we looked at revolved around sunspots. The data collected from sunspots seems random, which is a perfect test to see if ML can create a predictive model. After training the model based on the data set, we could watch the model get put to the test. While sunspots seem to be a random event, the ML model made predictions pretty accurately.
The sunspots were just the first randomness test, and ML could set itself so that it could see through the randomness. This process worked for sunspots, but what if we use a relatively random number generator package in python. The random python generator is pretty simple; there are better methods of creating randomness, but this is good enough for our testing. The ML model went to work on the data generated from python’s random function, which took much longer than predicting the sunspots. Once the model finished the test, the model, basically in its way, told us that this data was indeed random and could not predict. Our ML model that could predict sunspots could not foresee a pseudo-random series created by the random function in python. Oh, how the mighty fall. In my mind, I wonder if we could predict the random numbers that come out of a python random number generator. Many will say that the standard python random function is not ‘random’ enough.
Additionally, our neural network contained only three layers, with 30 nodes in the first layer, 10 in the second, and 1 in the final layer. Could a deeper neural network predict the number from this function more accurately? I am sure I could find the answer on google, but I think I might test this for myself later on as a project.
Today was a great first meeting, and I look forward to seeing where this group goes in the future.