The market for artificial intelligence is one of the fastest-growing markets in the World. The market value was estimated to be approximately $65.48 billion in 2020 and is expected to grow 38% before 2030.[Keshav K, 2022] Deep learning is a subcategory of artificial intelligence (AI) that is contemporary cutting-edge technol- ogy. Deep learning is a subclass of machine learning algorithms that uses multiple layers to progressively extract higher-level features from raw data. There are seven steps in building a deep learning model include: gathering data, preparing the data, choosing the model, training the model, evaluating, hyperparameter tuning, and pre- dicting. Understanding each step and how to build a good-performing model takes years of experience. That is where AutoML (automatic machine learning) can give the power to build one of these complex models to laymen and novices alike. AutoML automates every step in building a model besides gathering data. With basic knowl- edge of computers, anyone can train and deploy a complex predictive model. This will not only give novices the power of AI, but it will also free up the employees that are tasked with building and training models that could just be solved with AutoML. This will allow data scientists to focus more time and resources to the more complex problems. Companies that thrive in this domain can anticipate billions in revenue. In my paper ”Analysis and Survey of Auto Machine Learning (AutoML) for Computer Vision-based Gender Recognition”, I will research how AutoML works on a common but challenging problem in the computer vision space. Further I want to investigate if AutoML can deal with unbalanced data without creating biased predictions. With the growing problem of algorithm bias, AutoML will have to mitigate or remove these biases to make fair predictions. Algorithm bias is systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group, or class of data, of users over others. The questions that I want to answer are: How does AutoML compare with hand-tuned deep learning models in terms of performance? How sensitive to algorithm bias is AutoML? Is AutoML a solution to reduce or remove algorithm bias due to systemic racism?

Recommended Citation: Cook E., Ricanek K., Stoker G., Alamleh H., (2022). Analysis and survey of Auto Machine Learning (AutoML) for Computer Vision-based Gender Recognition. UNCW MS CSIS Proceedings. V. 16 , N. 0 .