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
This essay contrasts two stylized accounts of a quantitative trader’s working day. The less automated workflow involves checking strategies and orders manually, adjusting allocations and contracts, monitoring live positions, recording profit, slippage, and…
This article examines how the length of a training window affects an AI stock selection model. It compares long historical samples with shorter rolling samples and recommends checking each choice against a held-out backtest period. In the author's…
The article distinguishes high-frequency trading firms from quantitative investment firms by their typical origins, time horizons, capital models, and technical priorities. It describes high-frequency trading as focused on rapid processing and efficient…
This code reference presents several ways to measure dependence or distance between financial data vectors and matrices. It defines angular distance from Pearson correlation, plus absolute and squared variants that alter how negative or strong correlations…
The document introduces graph representations for modeling relationships among financial assets, arguing that stocks cannot be understood solely as independent observations. Graphs can encode different kinds of links, including sector membership, return…
This overview presents QLib as an open-source quantitative research platform that supports data preparation, model training, portfolio construction, risk analysis, backtesting, and execution research. It notes support for supervised learning, market-dynamics…
The document summarizes the ResNet paper’s approach to training deep convolutional neural networks. Instead of having stacked layers learn a full transformation directly, residual learning lets them learn a change relative to their input. The paper’s…
This code describes a deep learning approach for turning sequential market features into position signals. Its example model uses an LSTM layer followed by dropout and a time-distributed output constrained through a hyperbolic tangent activation. Training…
This document introduces generative adversarial networks (GANs) as a machine learning framework with two neural networks: a generator that creates samples and a discriminator that judges them. Their competing objectives form an adversarial training process.…
This report summary describes revisions to AlphaNet, a neural network that learns stock-selection factors from price and volume data. AlphaNet-v2 adds six ratio features, replaces pooling and fully connected layers with an LSTM to capture temporal patterns,…
The document explains how FinRL models automated stock trading as a Markov decision process. An agent observes market prices and features, acts in a simulated environment, receives rewards, and adjusts its policy to pursue higher cumulative reward. The…
This article explains how nonlinear feature representations can make complex classification possible with a linear decision boundary. It uses support vector machines as an analogy: mapping inputs into a higher-dimensional feature space can make patterns that…
The report describes applying Thompson Sampling, an online learning method, to adjust a multi-factor investment model as market leadership changes. It uses China A-share constituents from the CSI 300 and CSI 500, with valuation, profitability, growth,…
This article surveys Microsoft Research’s published work on AI for quantitative investing, especially Chinese equity selection. It explains methods that use graph neural networks to represent relationships among stocks, attention mechanisms to adapt to…
A user reports that a price-ratio buy condition never triggers during a year-long backtest, although the sell condition does. The response attributes the training error to StockRanker receiving only two factors, with a Boolean factor causing the model to…
This overview compares common machine learning methods by their assumptions, strengths, and limitations. It introduces the bias variance tradeoff and explains how model complexity can lead to underfitting or overfitting. It then surveys naive Bayes, logistic…
This short project note outlines a workflow for preparing data and fitting a predictive model. It divides SQL data construction into a base dataset, a normalized dataset, and a final feature-and-label stage. The note mentions cross-sectional factor ranking…
This document defines an abstract data-formatting interface for experiments using the Temporal Fusion Transformer (TFT). Dataset-specific formatters are expected to define column names and roles, fit scalers, transform features, reverse prediction…
This report introduces gcForest, or multi-grained cascade forest, as a tree-based deep learning model proposed as an alternative to deep neural networks. It argues that the method may suit financial data because it is described as more robust to…
This document reviews an empirical comparison of machine learning approaches for forecasting monthly returns in China’s A-share market. The described process trains models on prior data, uses a validation period for tuning, and tests predictions out of…
The document summarizes a study of machine learning models for predicting monthly Chinese A-share returns. It describes a rolling, annual retraining design that uses lagged stock and macroeconomic factors, with separate training, validation, and test…
This configuration describes a Qlib workflow for training a PyTorch feedforward neural network on China’s CSI 300 universe. It uses Alpha360 features, robust feature normalization, missing-value filling, and cross-sectional rank normalization of labels. The…
This notebook outlines a daily equity research workflow using CSI 300 constituents. It loads historical bars, builds an Alpha158 feature dataset, and defines training, validation, and test periods. The target is a forward VWAP return over a three-day…
The document discusses whether stochastic-process mathematics, including Brownian motion, still matters as machine learning becomes more common in quantitative finance. It presents several practitioner views: stochastic calculus has a stronger role in…