← All Collections
🧬 15 parts

Modern ML for Trading

The full gradient-boosting-to-neural-operator toolkit — ensembles, AutoML, probabilistic forecasting, distillation and pruning for latency, plus physics-informed architectures like Neural ODEs and Fourier operators.

  1. 01
    XGBoost for Return Direction: Class Imbalance and Decision Thresholds
    Aug 25, 2026 #machine-learning

    XGBoost for Return Direction: Class Imbalance and Decision Thresholds

    Return-direction classifiers are imbalanced problems, and 0.5 is the wrong decision threshold. Comparing scale_pos_weight, focal loss, and precision-constrained threshold optimization on crypto data — plus the engineering differences between XGBoost, LightGBM, and CatBoost.
  2. 02
    Ensemble Methods: Combining Weak Learners for Robust Alpha
    Aug 3, 2026 #machine-learning

    Ensemble Methods: Combining Weak Learners for Robust Alpha

    Three things about trading ensembles this blog has not covered: why the covariance term — not model count — dominates ensemble error, how stacking works as a signal-combination layer with out-of-fold meta-features, and whether turnover netting across many signals actually reduces execution cost.
  3. 03
    AutoML for Systematic Trading Pipelines
    Jul 30, 2026 #AutoML

    AutoML for Systematic Trading Pipelines

    Automated feature generation (tsfresh, Featuretools), budget-aware model search (FLAML's cost-frugal optimizer), and the WorldQuant formulaic alpha factory — the parts of the research pipeline this blog's search-and-overfit arc never covered, and what the arc's own results say about them.
  4. 04
    Multi-Task Learning for Simultaneous Price, Volume, and Volatility Prediction
    Aug 15, 2026 #deep-learning

    Multi-Task Learning for Simultaneous Price, Volume, and Volatility Prediction

    Does jointly predicting return, volume, and volatility actually help? Measuring loss-balancing schemes and diagnosing negative transfer through gradient cosine similarity — with a classical baseline and purged walk-forward folds.
  5. 05
    Anomaly Detection for Trading Bot Protection: From Z-Score to Transformer
    Feb 19, 2026 #algo trading

    Anomaly Detection for Trading Bot Protection: From Z-Score to Transformer

    Which anomaly detection methods actually work in crypto algo trading, how to build a cascading protection architecture, and why this is the foundation without which algo trading becomes gambling.
  6. 06
    Scoring Probabilistic Forecasts: CRPS, PIT Calibration, and DeepAR
    Aug 20, 2026 #forecasting

    Scoring Probabilistic Forecasts: CRPS, PIT Calibration, and DeepAR

    How to evaluate a predictive distribution honestly — CRPS as a proper scoring rule, the PIT histogram as a calibration diagnostic, and DeepAR sampling in GluonTS.
  7. 07
    Irregular Time in Tick Models: Continuous-Time Encodings vs. Plain Positional Embeddings
    Aug 22, 2026 #HFT

    Irregular Time in Tick Models: Continuous-Time Encodings vs. Plain Positional Embeddings

    Sequence models fed tick data still assume regular spacing. Three ways to tell a Transformer when a tick actually happened — learnable-timescale continuous encoding, ODE-RNN latent state, and delta_t as a plain feature — and the ablation that decides between them.
  8. 08
    Does a Full-Day Context Beat a Ten-Minute One? Flash Attention and the Sequence-Length Question
    Aug 4, 2026 #deep-learning

    Does a Full-Day Context Beat a Ten-Minute One? Flash Attention and the Sequence-Length Question

    Flash Attention makes a 23,400-step trading context computationally free — tiling, online softmax, and an IO bound of N^2 d^2 / M. Whether that longer context makes the model better is a separate question, and it has to be measured, not asserted.
  9. 09
    Knowledge Distillation: Compressing Trading Models for Low-Latency Deployment
    Aug 11, 2026 #model-compression

    Knowledge Distillation: Compressing Trading Models for Low-Latency Deployment

    The blog's standing answer to the accuracy-vs-latency tension is a two-stage fast/slow split. Distillation is a different answer: train one small model to mimic the ensemble. The KD loss, temperature, born-again nets, early exits for a variable latency budget, and the distill-to-FPGA pipeline — plus the measurements that would decide whether it beats the two-stage split.
  10. 10
    Model Pruning for Low-Latency Trading Inference
    Aug 14, 2026 #model-compression

    Model Pruning for Low-Latency Trading Inference

    Magnitude and structured pruning, the Lottery Ticket Hypothesis, movement pruning, distillation and 2:4 sparsity — the methods behind shrinking a trading model, and what still has to be measured before any of it ships.
  11. 11
    Gaussian Processes for Non-Parametric Price Modeling
    Aug 6, 2026 #bayesian

    Gaussian Processes for Non-Parametric Price Modeling

    Kernel design for financial time series — Matern roughness, locally periodic composition, spectral mixtures — and the marginal likelihood as a regularizer that needs no validation set. Plus the honest list of what still has to be measured before any of it is tradeable.
  12. 12
    Epistemic vs Aleatoric: Measuring What a Return Model Doesn't Know
    Jul 31, 2026 #bayesian

    Epistemic vs Aleatoric: Measuring What a Return Model Doesn't Know

    Every sizing rule on this blog treats uncertainty as one number. MC Dropout and deep ensembles split it into model ignorance and market noise — and those two deserve different position sizes.
  13. 13
    Neural ODEs: Does Continuous Time Beat a Delta-Time Feature?
    Aug 16, 2026 #deep-learning

    Neural ODEs: Does Continuous Time Beat a Delta-Time Feature?

    Neural ODEs, Neural SDEs and continuous normalizing flows for irregularly-sampled market data — and the one ablation that decides whether continuous dynamics are worth their solver cost.
  14. 14
    Hamiltonian Neural Networks: Does a Financial System Conserve Anything?
    Aug 8, 2026 #deep-learning

    Hamiltonian Neural Networks: Does a Financial System Conserve Anything?

    Hamiltonian Neural Networks are provably stable — but stability is worthless if the conserved quantity does not exist. Learning a scalar H via autograd, symplectic integration, and the falsification test that decides whether a financial (q, p) pair is canonical at all.
  15. 15
    Fourier Neural Operator for PDE-Based Financial Modeling
    Aug 5, 2026 #deep-learning

    Fourier Neural Operator for PDE-Based Financial Modeling

    Operator learning maps whole function spaces, not points — how the Fourier Neural Operator parameterizes a PDE solution operator in frequency space, what that buys for option pricing, and which of its claims still need measuring on real hardware.