Deep learning practice for high school student engagement in STEM careers

The use of deep learning activities at the educational stage is quite relevant for student engagement (positive or negative) in Science, Technology, Engineering, and Mathematics (STEM) careers. Physical contact with hardware, visual and interactive contact with software, and contact with algorithms are practices in which students observe the complexity of command could result in ingenious automation and task accomplishment (universal approximation theorem and/or probabilistic inference). Deep learning came to this work as a tool for solving a problem: How do we identify forage palm (Opuntia ficus-indica Mill, Cactaceae) for feeding goats, in an area under Semi-arid climate? High school students on advisory teachers of various disciplines (biology, chemical, physical, mathematics, geography, and others) attempt to answer to this question: Did student engagement happen by new context, problem to be solved, data collection (hardware), or data analysis (algorithm and software)? Sixty-seven high school students from Brazilian public schools (from 14 to 17 years old) participated in the proposed following teacher. High school students get more involved by practical and technological issues (mechanisms and hardware) than in scenario (new context) or by evaluation and discussion of the purpose. This data about the preference of one stage over other stages of Deep Learning determined the application for undergraduate courses in STEM. This was the most relevant data of this work, despite total engagement, students' preferences in one stage point out the student's academic vocation.