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 describes converting ADAM, originally a gradient-based optimizer used in neural networks, into a population-based method for general optimization problems. It reviews ADAM’s first- and second-moment updates, then outlines how the population…
The article introduces Kohonen maps, also called self-organizing maps, as an unsupervised method for mapping higher-dimensional observations onto a lower-dimensional grid while preserving neighborhood relationships. It explains competitive learning: each…
The article explains autoencoders as unsupervised neural networks that learn a compressed latent representation by encoding input data and training a decoder to reconstruct it. It outlines uses including dimensionality reduction, preprocessing, and noise…
The article ports a Python fractal feature set into an event-driven MQL5 engine while preserving causal timing. A fractal centered on bar c requires n later bars for confirmation, so the feature values are published at the confirmation bar and refer back to…
This article describes a workflow for tuning a two-hidden-layer deep neural network used to classify market data. It prepares pretraining, training, validation, and test sets; removes predictors judged statistically insignificant in earlier work; and uses…
This article presents reusable MQL5 utilities that extend matrix and vector operations, with an emphasis on preparing data for trading systems and machine-learning workflows. The functions described include reading numeric or categorical CSV data, writing…
The article extends an MQL5 ARIMA class with easier prediction methods and model saving and loading. It explains why forecasts need enough input observations: autoregressive terms require lagged values, non-contiguous lags preserve intervening time…
The article presents an architecture for sending live MetaTrader 5 feature vectors to a Python model service. Since MQL5 does not natively support gRPC’s HTTP/2 transport, the design uses a hand-written Protocol Buffers encoder and length-prefixed TCP…
This article builds an MQL5 library for the Matrix Profile, a time-series method that measures each fixed-length window against its nearest non-overlapping neighbor. Low distances identify recurring shapes called motifs; high distances identify unusual…
This article explains Segment, Shuffle, Stitch (S3), a learnable module for changing how time-series history is organized before a neural model processes it. S3 divides a sequence into non-overlapping segments, learns priorities that determine their new…
The article applies supervised learning to three Parabolic SAR and Relative Vigor Index signal patterns that previously failed forward-walk tests. It describes implementing the indicators in Python, connecting to MetaTrader price data, and training a neural…
The article introduces linear and Matérn covariance kernels for Gaussian Process Regression (GPR) and discusses their potential use with financial time series. A linear kernel represents simple linear relationships and is computationally light, making it a…
The article explains Beetle Swarm Optimization (BSO), a population-based optimizer that combines Particle Swarm Optimization (PSO) with Beetle Antennae Search (BAS). Each beetle probes objective values on either side of its position, then combines that local…
The article introduces Markov chains as probabilistic models for transitions among defined states. It explains the transition matrix and the memoryless assumption: the next state depends on the current state and elapsed time, rather than the full path taken…
This article explains a workflow for deploying a trained Torch time-series forecasting model in a MetaTrader 5 Expert Advisor using ONNX. It covers switching the model into inference mode, collecting and naming its inputs and outputs, exporting it, then…
The article presents nuclear norm maximization as an intrinsic reward for reinforcement learning agents exploring environments with sparse external rewards. It forms a matrix from encoded representations of a visited state and its nearest neighboring states,…
The article investigates whether planetary positions, lunar phases, solar activity, and planetary aspects can help forecast EUR/USD movements. It describes collecting daily astronomical data, aligning it with market history, and applying correlation analysis…
The article explains catch22, a compact set of time-series features selected from a much larger library for classification performance and low redundancy. Its features describe distribution shape, autocorrelation, nonlinear dependence, symbolic patterns,…
The article explains how triple-barrier labels can overlap in time, making financial machine learning observations dependent rather than independent. When many samples encode information from the same market interval, a model can effectively learn repeated…
This article argues that financial machine learning performance depends heavily on how the prediction target is defined. Unlike a fixed medical label, a trading target may represent returns, price appreciation, volatility, drawdown, or relative asset…
The article presents an inference-learning workflow for patterns built from moving averages and the stochastic oscillator. It frames inference as estimating latent structure from observed data, then describes a Wasserstein variational autoencoder that…
The article adapts a trend-scanning label method into a feature interface for machine-learning models. For each bar, it fits price against time across candidate window lengths and retains the window with the largest absolute slope t-statistic, along with…
The article explains Atomic Orbital Search (AOS), a population-based metaheuristic that maps candidate solutions to electrons distributed among imaginary orbital layers. Candidate fitness is treated as energy, with lower energy indicating a better solution.…
The article proposes using mutual information (MI) to choose which of several candidate trading targets a black-box regression model should learn. The examples compare future changes in price, moving averages, Stochastic, and RSI. MI is presented as…