Project Goals

• Faculty Member: Professor Brenda Rubenstein

• Course: Chem 1560: Accelerating Chemical Discovery

  • Equips students with data science and computational chemistry tools such as machine learning and neural networks to leverage their data for discovery.

  • Lab component includes Colab notebooks developed to teach data science concepts in the context of chemistry.

• Main Goal: Improve scaffolding of machine learning concepts within these notebooks.

  • Scaffolding: An instructional technique to move students progressively toward stronger understanding and independence in learning.

  • The instructor provides a simplified version of a lesson or activity and gradually increases complexity over time.

Methods

• Notebooks Written in Python Using sci-kit-learn

  • Introductory lessons to neural networks, utilizing resources from Brown CS department, DataCamp, and the internet.

  • Engaging exercises using sci-kit-learn to teach machine learning (e.g., regression, logistic regression, neural networks).

  • Examples:

    • Banknote Authentication: Predicts who threw which dart based on where it landed.

    • DeepChem examples: Predict binding energy of a protein-ligand complex, approximate wave functions.

  • Goal: Connect the dots, scaffold, and implement examples into the notebooks.

Challenges

• Lack of Prior Knowledge on Machine Learning Concepts

  • Self-learning via online and Brown CS resources to understand Python, machine learning, and neural networks conceptually.

  • DeepChem examples were technical and lacked guidance or annotation.

  • Difficulty in aligning examples with course content.

  • Worked with Professor Rubenstein to delve deeper into the code.

Successes

• Close collaboration with Professor Rubenstein to understand her vision and motivation.

• Successfully integrating examples into Colab notebooks, better supporting and guiding student learning.

• Gained insights into constructing course materials and understanding their purpose.

Reflections on Growth

• Consulting:

  • Consultant-client relationship, project management.

• Data Science:

  • Machine learning, neural networks.

• Learning and Teaching:

  • Active learning, scaffolding.

  • Effective communication through email and biweekly meetings.

  • Breaking down larger projects into smaller tasks with biweekly deliverables (sprints).

  • Understanding the role of machine learning in the broader data science field.

  • Using Colab notebooks as a medium for active learning (read and write code).

  • Starting with fun examples and incorporating chemistry examples for scaffolding.

Next Steps

• Continue to augment current notebooks, spending more time reviewing them.

• Verify that examples align with the methods used in the notebooks (e.g., sci-kit-learn vs. PyTorch vs. Keras).

• Continue collaborating with Professor Rubenstein to achieve her vision for the course and project.

Final Thoughts

• The project provided a great hands-on learning experience.

• Expressed gratitude to Professor Rubenstein and Professor Clark for their support and dedication.

• Appreciated the unique and rewarding nature of the course.