🧠 Daily Study Log [2025-07-14]
Another productive day — wrapped up key topics across TOEIC prep, CV experimentation, paper study, and research planning.
Finalized second phase of the EfficientNet paper review and pushed forward with model scaling experiments.

📘 TOEIC Study — RC Focus & Vocabulary Expansion

Focus:

• Practiced Reading Comprehension with a focus on inference-based questions
• Studied advanced vocabulary and contextual usage in long-form passages

Reflection:

• Improving speed and accuracy in passages with implicit reasoning
• High-level vocabulary retention through contextual immersion

🧪 Research Planning — Paper Study Bot Idea

Focus:

• Brainstormed a lightweight GPT-based paper review assistant
• Goal: internal use within the study group to summarize, comment, and track paper progress

Reflection:

• Tool could streamline group study and future publication planning
• Will scope out MVP features (summarization, annotation, retrieval) over the next few weeks

🧠 Paper Study — EfficientNet (Day 4 & 5)

Focus:

• Completed Section 4.3 and Table 2 analysis
• Wrapped up performance comparison across EfficientNet-B1 to B7
• Reviewed ablation studies, scaling methods, and performance trade-offs

Reflection:

• Compound scaling consistently outperforms naive scaling strategies
• Final summary coming up — will consolidate all notes into GitHub post soon

🧪 CV Experiment — EfficientNet Scaling Models (CIFAR-10)

Focus:

• Running comparative experiments on four scaling variants:
  • ✅ Base model
  • ✅ Depth-only scaled model
  • ✅ Width-only scaled model
  • ✅ Resolution-only scaled model
  • ✅ Compound-scaled model

Status:

• 🚧 Training still in progress — full results pending
• Metrics being logged: FLOPs, parameters, validation/test accuracy
• Summary table and plots will be generated post-training

Reflection:

• Expecting compound scaling to strike the best performance-efficiency trade-off
• Preliminary trends look promising — more insights to come after full runs

✅ TL;DR

📍 TOEIC: Focused on RC and advanced vocabulary — inference & context mastery
📍 Research: Ideated GPT-based study bot — paper summarization & annotation
📍 Paper: EfficientNet Day 4–5 — completed scaling analysis, wrap-up next
📍 CV: Training in progress for base/depth/width/resolution/compound models
📍 Tomorrow: Continue training, summarize EfficientNet, baseline for HAR comp