Machine learning and Semantic communication Researcher

QUANG THAI TRAN

Trần Quang Thái

Current Focus6G and AI applied Semantic Communication
Vault StructureProjects - Papers - Research - CV - Personal
Email: [email protected]GitHub: github.com/Fankos
""Where there's a will, there's a way""

Who am I?

This site is a structured home for

The

Research Highlights

Current themes and the parts of the vault that support them.

Machine Learning Systems

Implementation notes, repo walkthroughs, and systems experiments.

repo-analysisllm-servingsystems

Paper Reading

Structured notes for papers, equations, claims, and replication ideas.

kv-cacheloraefficient-training

Research Roadmaps

Open hypotheses, experimental plans, and checkpoints for future work.

research-roadmapexperiments

Personal Archive

Private memories, dates, and personal pages protected outside the public index.

privatecalendarmemories

Featured Projects

Representative entries with room for screenshots, diagrams, or project images later.

Recent News

Recent changes and public notes across the vault.

  1. ResearchEN
    Nonlinear Transform Source-Channel Coding for Semantic Communications

    NTSCC reframes DeepJSCC with learned compression priors, entropy-guided rate control, and a clearer source-channel tradeoff.

    semantic-communicationsjsccrate-distortionntscc
  2. ResearchEN
    Robust Information Bottleneck for Task-Oriented Communication with Digital Modulation

    A close read of DT-JSCC/RIB, where robustness comes from preserving just enough redundant task information under digital modulation.

    semantic-communicationstask-orienteddigital-modulationinformation-bottleneck
  3. ResearchEN
    Universal Joint Source-Channel Coding for Modulation-Agnostic Semantic Communication

    uJSCC is mainly a model-sharing argument for semantic communication across modulation orders, with switchable normalization doing the adaptation work.

    semantic-communicationsjsccmodulation-agnosticvector-quantization
  4. ResearchEN
    Vision Transformer-Based Semantic Communications With Importance-Aware Quantization

    ViT-IAQ uses pretrained attention as a bit-allocation signal, making semantic compression more modular but still dependent on calibration.

    semantic-communicationsvision-transformerquantizationattention
  5. ResearchEN
    Attention-Aware Semantic Communications for Collaborative Inference

    A practical edge-server inference scheme where a small ViT decides when to call the server and which patches are worth sending.

    semantic-communicationscollaborative-inferenceedge-aivision-transformer
  6. ResearchEN
    Markov-Enforced Discrete Diffusion Model for Digital Semantic Symbol Error Correction

    SSCDM treats noisy semantic symbols as a Markovian correction problem rather than a generic diffusion-denoising task.

    semantic-communicationsdiffusion-modelssymbol-error-correctionvector-quantization
  7. ProjectsEN
    Github Deep Learn

    A local-first research repo explorer that maps code structure, equations, and study sessions.

    repo-analysisresearch-toolspython
  8. ProjectsEN
    Personal Knowledge Vault

    The website itself: a low-cost archive for public work, paper notes, research logs, and protected personal pages.

    portfoliodigital-gardenastro

Selected Papers

Reading notes and research references.

  1. PapersEN
    CacheRoute: Prefix-aware Routing for LLM Serving

    Reading note stub for routing requests in ways that improve KV-cache reuse while balancing load.

    llm-servingsystemskv-cache
  2. PapersEN
    CARE-LoRA and Frozen-subspace Recovery

    Reading note stub for activation memory savings and trainable low-rank adaptation.

    loraefficient-trainingmemory