Artificial Market Simulations
The study suggests a new method for deep hedging in finance using artificial market simulations, which performs similarly to traditional models but has certain limitations.
6 sharesSource ↗
Quant LetterNo. 45
April 2024, Week 3
78 items across 9 sections, as sent to readers on 17 April 2024. Paper titles open their ML-Quant page; ↗ goes to the source.
- Items
- 78
- Papers
- 51
- Sections
- 9
- Issue
- No. 45
arXiv
Quantitative-finance and ML-for-finance preprints from arXiv.
8 items
Finance5
Risk Measure Derivatives
The paper provides the derivatives of any risk measures, including VaR and ES for portfolio loss variables, and presents asymptotic results for heavy-tailed portfolio loss variables.
3 sharesSource ↗
Informed Trading and Private Information
The paper investigates market-clearing equilibrium in a risky financial market, showing that insider welfare increases with signal precision and price impact can both benefit and harm traders.
3 shares1 citation todaySource ↗
Factor Risk
The paper presents factor risk measures to assess risk relative to major factors, discussing their use in regulatory capital requirement and risk-sharing issues.
3 shares2 citations todaySource ↗
PathIntegral Approximation
The paper improves the pricing of fixed income instruments within the Black-Karasinski model using neural networks, showing better results for multiple calibrations over extended periods.
2 shares1 citation todaySource ↗
Miscellaneous3
Backward Deep Learning for BSDEs
The study introduces a new deep learning algorithm for solving complex backward stochastic differential equations, proving its effectiveness with numerous numerical tests.
6 shares8 citations todaySource ↗
Weighted Moving Models
The paper outlines a method for approximating the exponentially weighted moving model using only a set number of past samples and convex optimization.
6 shares8 citations todaySource ↗
Japanese Financial LLM
The research focuses on developing a large language model specifically for Japanese finance, showing its enhanced performance on related benchmarks.
3 shares5 citations todaySource ↗
SSRN
Working papers in finance and economics from SSRN.
21 items
Quantitative11
Exposure Hedging Strategy
The paper presents a model for optimizing a dealer's hedging strategy in foreign exchange fixings, suggesting smaller exposures are fully hedged in the short term, while larger ones are hedged over a longer period.
73 sharesSource ↗
Implied Volatility in Defi Pools
The article introduces a breakeven implied volatility for decentralized finance pools, which aligns with a previous definition based on a market impact rule in traditional finance.
7 shares1 citation todaySource ↗
EPS Impact on Capital Structure
The study reveals that firms adjust their capital structures based on earnings per share (EPS) levels, with the impact of EPS becoming more significant after the Sarbanes-Oxley Act in 2002.
6 sharesSource ↗
Deep Learning for Cryptocurrency Trends
The research presents a new data preprocessing technique and a Convolutional Neural Networks (CNN) model for predicting Bitcoin market trends using 15-minute candlestick data.
2 shares1 citation todaySource ↗
Composite Likelihood Estimation for Gaussian Processes
The paper outlines a framework for composite likelihood inference of parametric continuous-time stationary Gaussian processes, focusing on the random log-spot variance of financial asset returns.
3 sharesSource ↗
ETFs Impact on Bond Liquidity
The article discusses how a bond's inclusion in a creation or redemption basket improves its liquidity, especially in the case of redemptions.
109 sharesSource ↗
Analyst Forecast
A machine learning method categorizes analysts' forecast revisions into five types, improving accuracy and reducing information asymmetry in earnings announcements.
2 sharesSource ↗
Interpolation Algorithm
A new trigonometric interpolation algorithm for even periodic functions, implementable via Fast Fourier Transform, optimizes operations and overcomes the classic algorithm's limitations.
3 sharesSource ↗
Stock Price Prediction with Deep Learning
The piece introduces a hybrid model that combines various AI techniques for predicting stock prices and optimizing trading decisions.
2 sharesSource ↗
Business Time Modeling for Commodity Markets
The piece presents a model that accurately represents commodity forward curves, useful for pricing exotic derivatives and managing commodity portfolios.
2 sharesSource ↗
Sentiment Impact
A new measure of anticipatory sentiment, created using statistical natural language processing, significantly influences macroeconomic and financial variables, including credit market stress indicators.
2 sharesSource ↗
Financial10
Risk Premia in Commodity Market
Machine learning methodologies reveal that momentum factors from equity, bonds, and currencies are priced into commodity returns, indicating a connection between commodity and other financial markets.
3 sharesSource ↗
Owner's Earnings & Stock Returns
Stocks with high owner's earnings tend to predict average stock returns and outperform other factors, providing significant alpha over the FamaFrench 6factor and q5 factor models.
14 sharesSource ↗
NASDAQ Electronic Trading Stock Dynamics
NASDAQ stocks exhibit a U-shaped pattern in bid-ask spreads and trading volumes due to aggressive trading at market open and close, especially for smaller stocks and those with larger order imbalances.
5 sharesSource ↗
Optimal Averaging for Portfolios
The article suggests a new method for optimizing portfolio weights by combining minimum-variance strategies, which enhances the variance and Sharpe ratio.
87 sharesSource ↗
Machine Learning in Fund Classification
The paper uses machine learning to categorize hedge funds, finding that those classified as systematic yield higher excess returns.
2 sharesSource ↗
Investment Sensitivity and Algos
The article reveals that algorithmic trading that supplies liquidity boosts firms' investment sensitivity to stock price and enhances operating performance, while the opposite is true for liquidity-demanding algorithmic trading.
2 sharesSource ↗
Cross-Momentum in Financial Markets
A study shows that equity futures and currency portfolios sorted by cross-momentum perform better than those sorted by normal momentum, especially in commodity exporting countries.
2 sharesSource ↗
Portfolio Choice with Genetic Programming
A new method for creating efficient portfolios using genetic programming and economic constraints has been developed, which doubles the out-of-sample Sharpe ratio of existing methods.
2 sharesSource ↗
Liquidity Provision in Futures
A study of the Chicago Mercantile Exchange's futures markets shows that aggressive trades and limit orders significantly contribute to price discovery, with most limit orders providing uninformed liquidity.
2 shares2 citations todaySource ↗
Downside Risk Estimation
A new model for estimating risk based on the corrected Cornish-Fisher expansion provides more accurate downside risk forecasts for various equity indices and commodity futures.
3 sharesSource ↗
RePEc
Economics working papers from RePEc's NEP field reports.
17 items
Finance5
AI in Finance: Data and Opportunities
Data and Opportunities: The article highlights the significant role of big data and AI in the finance industry, suggesting a blend of financial knowledge and data analytics for improved financial systems.
26 sharesSource ↗
Efficient Frontiers and Tangent Portfolio
The paper explores the characteristics of kinks in portfolio optimization, demonstrating their universal existence and the absence of tangency.
16 sharesSource ↗
Sectoral Volatility and Jump Risk
The study examines the structure of risk contagion across sectors, emphasizing the need for accurate identification of risk contagion structure for effective regulation.
14 sharesSource ↗
Memory-Enhanced Momentum in Futures
The research suggests a memory-enhanced momentum strategy for commodity futures markets, which surpasses traditional momentum in reward and risk, independent of the overall commodity market movement.
12 sharesSource ↗
Efficient Portfolios: Pareto-Dirichlet Approach
Pareto-Dirichlet Approach: The paper presents a Pareto–Dirichlet method to solve the MVSK portfolio optimization problem, enabling the creation of optimal portfolios efficiently.
11 sharesSource ↗
Statistical4
Predicting Systemic Financial Risk with Interpretable ML
Research suggests that machine learning models and financial stress index can accurately predict systemic financial risk, especially in stock and money markets.
26 sharesSource ↗
Innovative ML Workflow for China's Financial Crisis Prediction
A study outlines a method for predicting China's systemic financial crises using machine learning models and macroeconomic indicators, identifying six high-risk periods from 1990 to 2020.
15 sharesSource ↗
Gold Price Forecasting
The study suggests using the eXtreme Gradient Boosting machine learning model and Shapley additive explanations for precise prediction and understanding of gold price changes.
13 sharesSource ↗
Option-Implied Kurtosis
The research concludes that including risk-neutral volatility skewness and kurtosis in forecasting models does not improve their predictive power and may even lead to less accurate predictions.
11 sharesSource ↗
Machine Learning4
Predicting Output Trends in China
Machine learning study on Chinese data from 1993-2016 reveals credit is a better output predictor than money, but its effectiveness has lessened post-2007 due to financial development.
28 sharesSource ↗
Challenges in Reusing ML Applications
The article categorizes machine learning applications into four types based on reuse strategies and offers insights for their development and deployment.
23 sharesSource ↗
CostSensitive ML for Startup Investments
Machine learning models used to predict the success of Israeli startups can reduce investment risk, but may also limit potential profits by predicting fewer successful startups.
22 sharesSource ↗
ML for hierarchical time series forecasting
A multi-output regression model is proposed for better supply chain forecasting, using variables from different hierarchical levels to generate reliable predictions.
13 sharesSource ↗
Historical Trending4
FDI in Western Europe
The study uses machine learning to analyze factors affecting foreign direct investment in Western Europe, offering insights for capital allocation decisions.
24 sharesSource ↗
ML for Stock Market
The article suggests that machine learning could enhance returns on short-term investments in day-trading.
23 sharesSource ↗
Factorial Models in Economic Analysis
The article proposes new methods for studying time series and building factor models in response to changing trends in macroeconomic and sectoral modelling.
9 sharesSource ↗
Dragonfly Optimization Algorithm for Feature Selection
The Dragonfly algorithm, a Swarm Intelligence method, is being used to improve the classification of breast cancer.
7 sharesSource ↗
Machine learning
The general machine-learning papers the letter carried in 2023-25.
5 items
Recently Published2
Under-bagging Analysis
The under-bagging method improves classifier training from imbalanced data by enlarging the majority class, showing better performance than under-sampling and simple weighting methods.
18 shares4 citations todaySource ↗
RLHF Dataset Optimization
The DR-PO algorithm enhances Reinforcement Learning by incorporating offline preference data into online policy training, outperforming other techniques in summarization and the Anthropic Helpful Harmful dataset.
23 shares44 citations todaySource ↗
Historical Trending3
mu-Transfer: Neural Network Scaling Rules
Neural Network Scaling Rules: A study has found that the μ-Parameterization (μP) is generally effective in determining the best learning rates for large neural network models, although it doesn't work in all situations.
109 shares7 citations todaySource ↗
The Story of a Rashomon Quartet
A paper presents the idea of a Rashomon Quartet, four models with similar predictive performance but different data relationship explanations, emphasizing the need to visualize models beyond their performance metrics.
58 shares14 citations todaySource ↗
Neural Scaling Laws
A study examines a random feature model trained with gradient descent, providing insights into neural scaling laws, including the correlation between performance, training time, model size, and the increasing gap between training and test loss due to repeated data use.
69 shares112 citations todaySource ↗
Papers with code
Papers that shipped their code, from the Papers with Code feed (2023-25).
2 items
Rising2
LLMs as Regressors
The performance of large language models like Llama2, GPT4, Claude 3, etc., in linear and nonlinear regression is examined without any extra training or gradient updates.
75 shares
PolicyGuided Diffusion
The article presents a new method as an alternative to autoregressive offline world models, which allows for the controlled generation of synthetic training data.
42 shares
GitHub
Repositories the letter featured.
3 items
Finance3
Machine Learning Trading
The article shares the coding for the updated version of Machine Learning for Algorithmic Trading.
11,746 shares
Microsoft Qlib
Qlib is an AI platform that uses machine learning models for investment research and execution.
14,117 shares
Superalgos
The article introduces a free, open-source bot for automated trading of bitcoin and other cryptocurrencies.
3,846 shares
News
Industry news: funds, hiring, markets and regulation.
3 items
Quantitative3
Citadel Securities Tech Head Joins HFT Firm
Former Citadel Securities executive, Joshua Fisher, has joined high-frequency trading firm Hudson River Trading.
5 shares
Interactive Brokers Launches PB Service
Interactive Brokers has broadened its services for hedge funds, introducing a high touch prime brokerage and global outsourced trading services.
4 shares
Trader Accused of Selling Secret
Jane Street has adopted a unique and highly effective trading strategy.
1 shares
Podcasts
Episodes on markets, quant methods and economics.
9 items
Quantitative4
Pension Wars
Jeff Weniger discusses the potential effects of the pension wars concept on global equity markets and the role of financial engineering in investment strategies.
11 shares
EM Fixed Income
Jonny Goulden and Saad Siddiqui analyze the latest market developments and their impacts on the EM fixed income asset class.
8 shares
Signs of the Time
Rob Almeida and Bill Gevov discuss the future of interest rates, changing global dynamics, and the potential influence of AI on the investment community.
7 shares
Geopolitical Risk
Daniel Wagner talks about the complexities of the global geopolitical risk landscape and provides strategies for financial risk managers to better measure and mitigate geopolitical threats.
6 shares
Related5
Samara Cohen Insights
Barry Ritholtz of Bloomberg Radio interviews Samara Cohen from BlackRock Inc., discussing her career and roles at the company.
5 shares
Goldilocks Danger
LGIM's CIO Sonja Laud and experts discuss the potential impact of political and geopolitical factors on 2024's asset rally in their first official CIO call.
4 shares
Global Commodities Update
A JPMorgan podcast discusses the potential impact of a court ruling on future LNG demand and the effects of the USDA's April WASDE report on South American corn and soybean production.
4 shares
US Rates PostCPI
A JPMorgan podcast explores the impact of a high CPI print and strong jobs print on the 'high for long' narrative, and the reasons for the recent gold rally.
4 shares
Uranium Market Growth
MacroVoices hosts interview Justin Huhn, founder of Uranium Insider, discussing the uranium bull market and the future of the uranium mining sector.
4 shares
X / Twitter
Posts from quant researchers on X.
10 items
Quantitative6
Algorithmic Trading: Portfolio Construction
Portfolio Construction: Article 1: The lecture discusses the estimation of idiosyncratic covariance matrix, off-diagonal cluster analysis, and updating of short-term idiovol in portfolio construction.
9 shares
Machine Learning for Corporate Bond Return Prediction
The study finds machine learning models can predict corporate bond returns with significant accuracy, even after accounting for transaction costs.
4 shares
Stocks and High Inflation
The study shows that stocks and bonds perform poorly during high inflation periods, offering weak protection for investment portfolios.
3 shares
Financial History Analysis
The article analyzes the historical trends of stocks and bonds over the last 800 years using data from GlobalFinData.
2 shares
Equity Factors Debate
Frey's research suggests that around 40% of recorded equity factors are due to mispricing, with most factors indicating a return to fundamental values.
2 shares
Miscellaneous4
Tiny Time Mixers
Tiny Time Mixers (TTMs) provide quick pretrained models for forecasting multivariate time series with zero or few shots, with the pretraining process being efficient, taking only 36 hours using 6 A100 GPUs.
0 shares
Bookshelf Recs
The author suggests six books from their personal library for readers to enjoy.
0 shares
Linear Models Notes
The author commends Peng Ding's comprehensive lecture notes on Linear models and extensions from Berkeley.
0 shares
WILMOTT May 2024
The 2024 Wilmott magazine edition contains exclusive articles from renowned columnists, educators, and researchers.
0 shares