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 overview organizes machine learning methods in two ways: by how they learn from data and by similarities in their model structures. It distinguishes supervised learning with labeled outcomes, unsupervised learning that finds structure in unlabeled data,…
This article introduces Transformer attention and adapts the architecture to stock-factor time series. Its proposed model adds Time2Vec representations to factor inputs so the network can represent both periodic and non-periodic time patterns. Self-attention…
This strategy note outlines a Chinese equity model that ranks stocks using three cross-sectional features: market capitalization, turnover, and five-day price change. Each feature is converted to a daily percentile rank. A shallow XGBoost regressor is…
This brief discussion proposes a feedback loop for a short-horizon stock strategy: observe the result of a trade, such as a buy today and sell tomorrow, then use the realized daily return to adjust strategy parameters. The intended outcome is a model that…
This configuration lays out a Qlib workflow for training a Transformer model on Alpha360 features and ranking CSI 300 stocks. It specifies Chinese market data, a close-to-close forward return label, robust feature normalization, missing-value filling, and…
This Chinese-language note points to a reproduction of a China International Capital Corporation research report on using intraday trading data to give an early warning of crowded institutional trading. It directs readers to a code-share implementation and…
This report compares artificial intelligence stock-selection portfolios across five Chinese equity universes and benchmark or sector-neutral setups. The models include support vector machines, random forests, naive Bayes, and logistic regression. It reports…
This research summary compares machine-learning stock-selection models trained on conventional return labels with models trained on alternative labels based on information ratio and Calmar ratio. The rationale is that these labels capture more of the path…
The document proposes combining two stock-ranking models in a stock-picking strategy: one predicts stocks likely to rise, and another predicts stocks likely to fall. The intended process is to use the downside model to remove candidates flagged as likely…
This overview traces supervised learning methods from linear discriminant analysis and naive Bayes through perceptrons, logistic regression, nearest neighbors, backpropagation, boosting, support vector machines, random forests, and distance metric learning.…
This report summary develops a case for timing Chinese bank stocks using macroeconomic indicators. It frames banks as cyclical and links their earnings and net interest income to economic activity, credit-related asset growth, and interest margins. The…
This page collects questions for a quantitative trading discussion rather than supplying answers. Topics include the Sharpe ratio, overfitting and the gap between backtests and live results; outlier treatment and avoiding look-ahead bias when building…
This Q&A explains several BigQuant modeling interfaces. It distinguishes a DataFrame, the tabular data type used by pandas, from a DataSource, a wrapper for data access and transfer between modules that can hide whether storage is in memory, local files, or…
This workflow demonstrates an end-to-end daily equity modeling process using CSI 300 constituents and vn.py’s AlphaLab tools. It loads constituent histories, builds an Alpha158 dataset, and divides observations into training, validation, and test periods.…
The post outlines a Chinese equity screening idea focused on metaverse-related stocks. It combines three filters: membership in the selected industry, a positive institutional activity measure, and a threshold condition involving price movement and…
This expert advisor aims to trade range-bound conditions in EURUSD and GBPUSD. An ONNX model classifies the market as sideways, bullish, or bearish using nine inputs, including EMA200 slope, distance from the average, ATR, candle structure, breakout…
A user reports that every label in a trading demo appears to be zero. The response explains that the platform’s visualization shows only a subset of the data, and the displayed rows may all happen to carry the zero label. It recommends inspecting the full…
The document outlines why quantitative investors use machine learning to analyze growing financial datasets, forecast market movements, and help select portfolios. It describes machine learning's capacity to represent nonlinear relationships and process data…
This meetup Q&A covers several practical quantitative trading topics. It explains rolling training as repeatedly updating a model using a moving or expanding historical window, and distinguishes incremental updates from retraining on the full dataset. It…
The document summarizes a research paper that uses a neural network to produce portfolio weights for exchange-traded funds. Instead of forecasting expected returns and then feeding those forecasts into a conventional optimizer, the approach directly tunes…
This Chinese-language seminar summary frames AI adoption as a response to falling asset-management margins despite growing assets under management. It sketches a progression from adjusting active portfolios using factor analysis, to building long-short…
This study outlines a Deep Q Network approach to timing trades in one stock. Instead of storing action values for every possible market state in a table, a neural network estimates values for available actions. The agent stores state, action, reward, and…
This introduction explains the Skip-gram model as a method for predicting nearby context words from a chosen center word. It uses a short example sentence and a window extending two words to either side. The explanation defines one-hot representations, dense…
This article introduces one-dimensional convolutional neural networks as a way to process financial sequences. A convolution applies shared filters across local time windows, allowing the model to learn patterns that can appear at different points in a…