Grace Hopper Celebration 2025 Presentation
Demystifying the mathematics powering deep learning

Deep learning often feels like magic — systems that can recognize faces, translate languages, and generate human-like text. But behind that magic lies mathematics — the true engine driving every neural network.

This repository contains the presentation materials from my talk at Grace Hopper Celebration 2025, where I demystify the math that powers deep learning and show how concepts from linear algebra, calculus, probability, and optimization come together to make AI work.

Through intuitive visuals and real-world examples, this presentation explores:

• Linear Algebra: How vectors and matrices move data through neural network layers
• Calculus: How derivatives and gradients help models learn through backpropagation
• Probability: How probabilistic reasoning guides predictions and uncertainty quantification
• Neural Networks: Putting it all together to understand how deep learning systems function

This is not about solving equations — it's about understanding the "why" behind the algorithms.

├── presentation/ │ └── GHC2025_Math_Behind_Deep_Learning.pptx ├── sections/ │ ├── 01_linear_algebra.pdf │ ├── 02_neural_networks.pdf │ ├── 03_calculus.pdf │ └── 04_probability.pdf ├── resources/ │ └── references.md └── README.md 

Attendees will walk away with:

1. Conceptual Understanding: A clear grasp of how mathematics shapes model behavior
2. Practical Confidence: The ability to interpret AI systems more effectively
3. Actionable Insights: Knowledge for building smarter, fairer, and more explainable AI solutions

• Vectors and matrices as data representations
• Matrix multiplication in forward propagation
• Dimensionality and transformations
• Weight matrices and their role in learning

• Architecture fundamentals
• Forward propagation
• Activation functions
• Layer compositions and deep networks

• Derivatives and gradients
• Chain rule in backpropagation
• Gradient descent optimization
• Learning rates and convergence

• Probabilistic predictions
• Loss functions and likelihood
• Uncertainty quantification
• Bayesian perspectives in deep learning

Aditya Hajare
Senior Software Architect, Enterprise Innovation - AI/ML
Fannie Mae

• 🎓 IEEE Senior Member | AI Policy Committee 2025
• 🚀 Co-Founder & CTO, Infinict (Web3 Fintech)
• 📚 4 Patents | 15+ Publications
• 🌍 Python Software Foundation Member
• 💡 Passionate about education equity and AI explainability

Connect with me:

• LinkedIn: https://www.linkedin.com/in/itis-aditya-mehra/

Grace Hopper Celebration 2025
The world's largest gathering of women and non-binary technologists

Session: The Math Behind the Magic: Understanding the Role of Mathematics in Deep Learning
Date: November 6 2025 Location: Chicago

For those interested in diving deeper:

• Deep Learning Book by Goodfellow et al.
• 3Blue1Brown Neural Networks Series
• Mathematics for Machine Learning
• Anthropic's Prompt Engineering Documentation

Found an error or have suggestions for improvement? Feel free to:

• Open an issue
• Submit a pull request
• Reach out directly

This repository is licensed under the MIT License - see the LICENSE file for details.

Special thanks to:

• Grace Hopper Celebration organizing committee
• The AI/ML community for continuous inspiration
• Everyone working to make AI more accessible and understandable

For questions, collaboration opportunities, or speaking engagements:

• LinkedIn: [(https://www.linkedin.com/in/itis-aditya-mehra/)]

⭐ If you find this helpful, please consider starring this repository!

Last Updated: November 2025