Current Module: 03 (Traditional Machine Learning & Time-Series Data)

Live Learning Log

A chronological terminal of my journey through the AI VIET NAM (AIO 2026) Intensive AI/ML Program.

sync
IN PROGRESS08/2026 - Present (Month 3)

[MODULE_03] Traditional Machine Learning & Time-Series Data

Deep dive into tree-based algorithms (Decision Trees, Random Forest, AdaBoost, Gradient Boosting, XGBoost, LightGBM) and Time-Series Data Analysis (Aggregation, Decomposition, Medical & Energy Analytics).

construction

Research Media & Visualizers
Will Be Published Upon Completion

>Dataset: To Be Determined...
>Pipeline: Tree-based Models & Time-Series Forecasting
>Validation: Pending Assignments...
>Status: Currently studying concepts and working on assignments...
lock_clockInteractive Demolock_clockGitHub Repositorylock_clockYouTube Demo
check
COMPLETED07/2026 - 08/2026 (Month 2)

[MODULE_02] Player Style Clustering from Football Event Data (Research Stream)

Unsupervised behavioral profiling of 1,016 players across 1,140 matches in Premier League, La Liga, and Serie A (2015/16). Benchmarking feature scalers, K-Means (k=5), and hierarchical tactical sub-clustering.

>Pipeline: RobustScaler + K-Means (k=5) + Hierarchical Sub-Clustering
>Validation: Purity: 69.7% | NMI: 0.546 | KNN 5-fold CV: 96.6% ± 0.8%
>Status: All code, modeling, interactive visualizers, YouTube video demo & verified certificate published
check
COMPLETED06/2026 - 07/2026 (Month 1)

[MODULE_01] Multimodal Similarity Matching for Images and Texts (Research Stream)

Implemented and benchmarked multiple image-text similarity matching approaches using ResNet50 baseline, Least-Squares projection matrix (W), and OpenAI CLIP contrastive multimodal embeddings.

>Dataset: Flickr8k (8,000 images, 40,000 paired text captions)
>Pipeline: ResNet50 + TF-IDF vs. Least-Squares Projection vs. CLIP (ViT-B/32)
>Validation: Mean Reciprocal Rank (MRR), Recall@1, Recall@5, Recall@10
>Status: All code, report, verified certificate & YouTube video demo published