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 examines how large datasets and complex models can make factor discovery more productive while increasing the risk of overfitting. It contrasts hypothesis-led research, where an economic or behavioral rationale is tested, with data-led mining,…
This tutorial explains attention and Transformer components before outlining a time-series stock ranking model. It adds Time2Vec features to factor sequences so the model can represent both periodic and linear time patterns, then applies self-attention,…
This brief guide explains how to exclude delisted stocks from both the training and validation data flows in a visual AI stock-selection strategy. It instructs users to insert a shared “remove delisted stocks” module before the missing-data handling module…
This preliminary example applies a functional LSTM network to Chinese equity data, using the CSI 300 as its prediction target. The sequence input contains six market features—closing, opening, high and low prices, amount, and volume—across 30 time steps. The…
This introductory overview explains how neural networks use layers of connected units to model nonlinear relationships. It sketches the roles of input, hidden, and output layers, activation functions, and deeper architectures such as convolutional and…
This configuration defines a Qlib workflow that trains a CatBoost model on Alpha360 features for CSI 300 constituents. It uses Chinese market data from 2008 through mid-2020, with training through 2014, validation over 2015–2016, and testing from 2017…
This tutorial presents a FinRL workflow for applying deep reinforcement learning to a portfolio of Dow Jones stocks. It formulates trading as a Markov decision process: the agent observes market features and holdings, chooses buy, sell, or hold actions, and…
This FinRL tutorial outlines a deep reinforcement learning workflow for trading a basket of Dow Jones stocks. It models trading as a Markov decision process: the state contains market and indicator data, actions specify share purchases or sales, and rewards…
The document explains how genetic programming can evolve formulas from stock price and volume inputs to discover candidate equity selection factors. It outlines the evolutionary cycle of selection, crossover, and mutations, and describes adapting a genetic…
The document surveys proposed Transformer applications in quantitative finance, including time-series forecasting, factor analysis, risk estimation, portfolio selection, execution, order-book modeling, and market-manipulation detection. It attributes their…
The report describes a stock-selection approach that forecasts the coming information coefficient of seven style factors, then adjusts their weights according to expected effectiveness. Its features include each factor’s historical information coefficient…
This Python class appends expanded values to a tabular dataset using polynomial bases or feature products. Its available polynomial families are Chebyshev, Legendre, Laguerre, and ordinary powers; the requested degree controls how many orders are generated.…
This configuration describes a Qlib workflow for training a LightGBM model and using its forecasts in an enhanced-indexing portfolio strategy. It specifies CSI 300 instruments and Shanghai’s CSI 300 index as the benchmark, with Alpha158 features and…
This data-preparation module builds time-series inputs for a deep-learning momentum model. It reads close prices, clips extreme values using an exponentially weighted mean and standard deviation, then derives daily returns and volatility. The target is a…
This support thread describes a BigQuant stock-ranking workflow that backtests successfully but fails when launched as simulated trading. The logs show that feature extraction runs, but later extraction steps return empty datasets; the ranker then receives…
The article defines market sentiment as participants’ collective outlook, which may be optimistic, pessimistic, or neutral and can be shaped by economic releases, company results, politics, news, and social discussion. It introduces sentiment analysis as a…
The document proposes a Chinese stock selection rule combining price activity and large-order flow. It screens for stocks with an amplitude above a stated threshold, excludes stocks that hit the upper limit on the previous day, and selects when the product…
This review describes a Bayesian dynamic panel model for evaluating US mutual fund performance and its persistence. The model allows fund returns to have time-varying variance and covariance, cross-fund dependence, serial correlation, latent volatility, and…
The code describes a manager for tuning model hyperparameters through random search over discrete parameter ranges. For each trial it samples one value per tunable parameter, adds fixed experiment parameters, checks that the parameter set is complete, and…
This document summarizes several modeling choices for quantitative trading. It argues that machine learning may be easier to apply in high frequency strategies than in lower frequency settings, and that nonlinear models can capture more from data than linear…
This discussion concerns adapting a stock-ranking template to use financial factors in a machine-learning strategy. The author describes a forward-return label based on buying at the next day’s open and selling at the close five trading days later, then asks…
The document describes a compact artificial neural network with one hidden layer, trained using backpropagation, and a chart indicator that visualizes price movement while projecting future closes. The indicator uses recent open prices as inputs, compares…
The document poses a model validation question rather than giving a complete answer: after training a multi-factor model on historical data and evaluating it on a later year, should a trader deploy that fitted model or retrain it on a newer rolling window?…
The document explains softmax as a transformation from a vector of scores to nonnegative values that sum to one, making them interpretable as a probability distribution. It describes its role in the continuous bag-of-words model in Word2Vec, where model…