Speaker Event – 1: Autonomous Object Detection
With autonomous vehicles taking off in the past several years, it is good to explore the aspect that is instrumental in creating a good autonomous vehicle system: Object Detection. The vehicles’ cameras are feeding the autonomous system what objects it is observing (traffic lights, traffic signs, and other vehicles). The autonomous system then utilizes the information to make a decision to turn left, go straight, and etc. Let’s create an Object Detector model using Darknet and YOLOv4 to see if we can emulate the “eyes of the car”.

Speaker Bio:
Ridwan Alam is a self-taught software developer, and recently completed the Metis Data Science bootcamp. He enjoys working with Computer Vision, NLP, Machine Learning, along with working on projects in React to help create stimulating visuals. He takes a key interest in emerging technologies such as Hyperloop and Bitcoin as he sees these tools as pathways into our future. Please reach out to him if you would like to talk more about Data Science, Web Development, or anything tech! If you are aware of any Data Science opportunities please contact him as he is actively searching since his Data Science bootcamp!

Speaker Event – 2: Understanding SHAP values for better model interpretation
Although there have been sweeping improvements in the predictive power of machine-learning models, the interpretation of these models and their results have become more complex. While a single decision tree or basic regression model can easily be interpreted and the importance of features determined, more complicated ensembles or deep-learning models are not as readily understood. SHapley Additive exPlanation (SHAP) values provide an agnostic way of exploring how different features impact model results. Although this concept was originally devised for game theory, it has also enjoyed great success with machine learning. In this talk I will explore SHAP values, how one calculates them and how to apply them to feature importance in machine learning.

Speaker Bio:
Dan has a PhD in Physics & Astronomy from Johns Hopkins University and has worked to apply machine learning techniques to Cosmology. After exploring the mysteries of the Universe, he transitioned to data science full-time. Dan now analyzes insurance claims from the healthcare industry to find undiagnosed patients with rare diseases and explores patient journeys through different disease states.

source