Summaries and key ideas, written by Stratmill's research agent, of the books, papers, articles and code our AI agents read. Each page links to its original.
5,018 documents
The article explains Linear Discriminant Analysis (LDA) as a supervised method for reducing feature dimensions while retaining information that separates labeled classes. It describes calculating within-class and between-class scatter matrices, solving an…
The article represents trading histories as sequences of outcomes and applies probability models to describe them. It first reduces trades to an equity curve, arguing that different trade arrangements can be equivalent when they produce the same curve. A…
The article surveys classification estimators in Scikit-learn and explains a workflow for exporting supported models to ONNX for inference in MQL5. It uses Fisher’s Iris dataset, whose samples have four measurements and three species classes, as a common…
This article outlines a proposed trading system that represents instruments and market variables as a network, applies causal discovery, and uses a Vector Autoregression model to forecast multiple related time series. It describes FCI as the causal discovery…
This article proposes changes to Atomic Orbital Search (AOS), a population-based method for optimizing objective functions. In the original method, particles are grouped into layers whose average positions and energies influence movement. The modification…
The document introduces a multi-task learning design that uses ResNeXt to extract features from financial data and feed several prediction tasks. It explains grouped convolutions, residual connections, bottleneck blocks, and cardinality, then describes a…
The article adapts Craig Reynolds’ Boids flocking model as a population-based optimization method. Agents move according to separation, alignment, and cohesion, with each rule controlled by distance and weight parameters. The described implementation also…
The article describes Prioritized Experience Replay (PER) for off-policy reinforcement learning, focusing on its use with Soft Actor-Critic. Instead of sampling past transitions uniformly, PER assigns priorities from temporal-difference errors so experiences…
The document adapts ideas from a graph transformer generative model, originally developed for architectural layouts, to trading-agent representation learning. Its encoder combines transformer attention, which can represent broad relationships, with graph…
The document describes a particle swarm optimization (PSO) method for searching Expert Advisor parameter sets in MetaTrader 5. Each particle has a position, velocity, and personal best; particles are grouped so they can also use a group best. Their positions…
This article presents a data preparation workflow for deep neural network experiments on financial time series. It starts with terminal OHLCV quotes, organizes them by time, and derives price series and technical predictors, including adaptive trend filters,…
The article explains temporal-difference (TD) learning as an incremental reinforcement-learning method that updates a state’s value using the immediate reward and the estimated value of the next state. It contrasts this state-value approach with SARSA’s…
This document explains ways to connect MetaTrader 5 with MATLAB for mathematical modeling and trading-system development. It describes how to map MQL5 numeric types to MATLAB arrays, handle strings with different encodings, and account for differences in…
The article proposes turning lunar and solar timing into features for a currency forecasting model. It describes calculating phase angles from fixed cycle lengths and currency-specific reference dates, encoding phases with sine and cosine harmonics, and…
The article introduces Dichotomy of Control (DoC), a reinforcement learning approach that separates effects a policy can influence from environmental randomness. It explains the motivation through Decision Transformers: conditioning an agent on a desired…
The article presents Behavior-Guided Actor-Critic (BAC) as an alternative to entropy-regularized Soft Actor-Critic for reinforcement learning with continuous actions. BAC uses an autoencoder trained on state-action pairs: its reconstruction error serves as a…
This article describes a pipeline for moving historical MetaTrader 5 bars into a Python service as preparation for machine-learning analysis of price spikes. Its implementation retrieves OHLC history, serializes timestamps and selected prices into JSON, and…
The article explains CATCH, a neural network framework for detecting anomalies in multivariate market time series. It transforms normalized data with an FFT, divides the real and imaginary spectra into patches, and uses a channel-masked Transformer to learn…
The article explains GinAR, an end-to-end neural architecture for forecasting multivariate time series when some variables have incomplete or entirely missing histories. It combines a simple recurrent unit with Interpolation Attention, which reconstructs…
This article adapts the Diversity Is All You Need approach to training trading agents. Its first phase learns a set of skills without an external task reward: a skill-conditioned agent acts from the observed market and account state, while a discriminator…
The article introduces neural network architectures and learning approaches, then focuses on deep networks for classifying financial market data. It describes autoencoders and restricted Boltzmann machines as building blocks for stacked models, and outlines…
The document outlines an experimental EURUSD workflow that trains a CNN–LSTM model on scaled closing prices and exports it to ONNX for use in MetaTrader 5. It describes a historical-data split for training and evaluation, dropout and early stopping during…
The article concludes a series implementing HimNet, a neural framework for spatiotemporal forecasting, and focuses on its temporal encoder. The architecture uses parallel processing for different dependency types and forms temporal embeddings at two scales.…
The article presents a modified Spiral Dynamics Optimization method for continuous optimization in multiple dimensions. It describes limitations of pairing coordinates to form spirals, especially for odd-dimensional problems and functions whose coordinates…