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 PSformer, a Transformer architecture for multivariate time series forecasting. It divides input series into patches, groups patches at matching positions across variables into segments, and applies attention to model relationships across…
The article frames stop-outs as a problem of price moving against a correct directional view before eventually reversing. It argues that simply widening stops creates inconsistent risk, while waiting for confirmation may only delay a stop-out. Its baseline…
The article describes the Dingo Optimization Algorithm, a population-based method for searching an optimization space. Each candidate solution is represented as a member of a population, and its position is updated using modeled group attacks, individual…
This article motivates Advantage Actor-Critic by contrasting value-based Q-learning with policy-gradient methods. Q-learning estimates action values and can support stepwise decisions, but its training can be biased and requires exploration choices. Policy…
This article introduces CatBoost, a gradient-boosted decision tree library that can process categorical features, and describes its ordered encoding and symmetric tree structure. It compares CatBoost with LightGBM and XGBoost, noting tradeoffs in…
The article explains MacroHFT, a reinforcement-learning framework for high-frequency cryptocurrency trading. Its original design classifies fixed-length market blocks by trend and volatility, then trains specialized agents for the resulting market regimes. A…
The article introduces seasonal decomposition as a way to separate a time series into trend, seasonal, and residual components. It outlines additive and multiplicative forms: trend is estimated with a moving average centered to align with the original…
The article introduces Gaussian processes as a statistical forecasting framework that predicts a distribution, including a mean and uncertainty, rather than only a point estimate. It focuses on the radial basis function kernel, which measures similarity…
The article explains Adam-mini, an Adam optimizer variant designed to reduce the memory used for per-parameter second-moment estimates. It groups model parameters into blocks and tracks a moving average of squared gradients for each block, using the block…
The article describes Elite Crystal Evolution, a population-based method adapted from the Crystal Energy Optimizer for continuous search. It ranks candidate solutions by fitness, treating the top ten crystals in a population of fifty as elites. Elites refine…
This article develops Symbolic Fourier Approximation (SFA), a way to encode normalized price windows as short symbolic words for similarity search. It replaces SAX’s piecewise-average representation with selected low-frequency Fourier coefficients, then…
This article completes an MQL5 Gaussian process library by describing interfaces for covariance kernels, likelihoods, and inference. It details RBF, linear, and periodic kernels, including their hyperparameters and covariance derivatives, then applies the…
The article presents Approximate Entropy (ApEn) as a measure of recurring structure in rolling sequences of closed-bar log returns. It explains how the calculation embeds returns into short pattern vectors, counts similar patterns using Chebyshev distance…
This article describes connecting MetaTrader 5’s standard signal classes to the CStrategy trading engine so that strategies can reuse existing indicator-based patterns. It surveys the library’s class structure, including signal modules, indicators, price…
This article introduces support vector machines as supervised models for classification and explains how their decision boundary is chosen. It describes hyperplanes, support vectors, and the maximum-margin objective, then distinguishes a linear model’s…
This article evaluates whether oil and gold prices improve forecasts of USDZAR compared with the pair’s own OHLC data. It describes exporting price histories, reversing their order into chronological sequence, aligning observations by timestamp, and…
The article explains how an MQL5 application can communicate with external quantitative software using raw TCP sockets and a native implementation of the ZeroMQ Message Transfer Protocol. It describes the request-reply pattern and the roles of protocol…
The article describes HimNet, a neural forecasting framework designed to account for differences across instruments, venues, and time periods. It represents time of day, day of week, and each series or location with trainable embeddings. Similar contexts can…
The article proposes detecting scale-varying symmetric patterns in price data through sliding-window correlation. For each starting point, it compares the first half of a window with the reversed, sign-inverted second half across a configured range of even…
The article describes a pipeline that turns price changes into binary sequences, groups the bits, maps them to BIP39 words, and feeds word indices to a transformer encoder. The stated aim is to predict subsequent market sequences and inspect word patterns…
The article describes Enhanced Colliding Bodies Optimization (ECBO), a population-based metaheuristic for minimizing an objective function. Candidate solutions are treated as bodies with masses based on solution quality. After ranking the population, the…
The article explains how a single-neuron perceptron and a multilayer perceptron produce predictions, then describes how training adjusts their weights. A perceptron combines inputs with weights and a bias, applies an activation function, and can classify…
The article explains the Multi-future Transformer, a neural architecture for predicting multiple plausible futures for several interacting agents. It frames future uncertainty as a mixture of unimodal scene outcomes, separating distinct interaction modes…
The article explains Bayesian Online Change-Point Detection (BOCPD), an online method that updates each bar using only current and past observations. It maintains a probability distribution over run length, the time since the latest regime break, and uses…