Research
Review notes on time-series forecasting models and trading signals.
A trading research note comparing TD3+BC, ReBRAC, CQL, IQL, Decision Transformer, and diffusion-based offline RL.
A JeTech research note on D4RL, dataset quality, normalized scores, Minari, and what trading research should borrow from the benchmark.
A roadmap connecting market world models, offline RL algorithms, time-series representation learning, dataset generation, and real market gates.
A JeTech research note defining the world model as a synthetic market path generator rather than an offline RL algorithm.
Root-cause notes for low candidate yield and the updated missing-combo learning-quality loop.
Why synthetic-data RL training and real five-year backtest gates diverge for AI trading models, including TradeFi, crypto expansion, and rolling Sharpe.
Runtime preset and registry gate notes for the first JeTech iTransformer v1 candidate model family.
FOIL 논문의 핵심 내용을 요약하고, 다양한 시계열 예측 모델에 적용한 방법론과 구현 사례를 상세히 분석합니다.
지난번 소개한 AutoTimes 모델을 실제 데이터로 학습시켜 트레이딩 성과를 살펴보겠습니다.