🧠 Daily Study Log [2025-09-17]
Learned statistical model comparison in EDA, explored full-cycle big data practices, studied Havruta-style AI discussions, proposed an IoT-connected alarm idea, and continued diffusion model paper review (forward process).

📚 Classes

• Big Data Analysis & AI Modeling: Learned how to perform EDA based on statistics, compare multiple models statistically, and apply interpretation methodologies for those models.
• Big Data Practical Analysis: Covered the end-to-end workflow in enterprise settings, from data importing to storage strategies, mirroring real-world corporate practices.
• AI-based Havruta Education: Explored the Havruta (Jewish dialogic learning) method and practiced AI-assisted discussions to deepen understanding and generate new perspectives.

💡 Idea

• Proposed a new lifestyle idea: IoT Connected Alarm
• Motivation: Difficulty in waking up despite existing footpad alarms.
• Concept: Upgrade traditional alarm platforms with IoT connectivity, integrating with other devices to enhance wake-up effectiveness.

📖 Paper Study

• Continued study of Denoising Diffusion Probabilistic Models (DDPM).
• Day 2 – Forward Process (Noising)
  • The forward process gradually corrupts data with Gaussian noise in a Markov chain until it becomes pure noise.
  • Key properties: no learning required, direct access to noisy samples, guaranteed convergence to Gaussian noise.
  • Intuition: like repeatedly adding static to a photo until all original content disappears.
  • Insight: This provides a clear blueprint for the reverse process, giving the model a structured task to learn.

✅ TL;DR

📍 Learned statistical model comparison & interpretation in EDA
📍 Covered full enterprise workflow of big data analysis
📍 Studied Havruta-style education with AI integration
📍 Proposed IoT-connected alarm lifestyle idea
📍 Reviewed forward process in DDPM (systematic noising)