Neural Networks for Finance
The study investigates the use of supervised autoencoders in improving financial forecasting through precise parameter tuning.
3 shares27 citations todaySource ↗
Quant LetterNo. 43
April 2024, Week 1
65 items across 6 sections, as sent to readers on 3 April 2024. Paper titles open their ML-Quant page; ↗ goes to the source.
- Items
- 65
- Papers
- 45
- Sections
- 6
- Issue
- No. 43
arXiv
Quantitative-finance and ML-for-finance preprints from arXiv.
5 items
Finance5
RL Agents in Market Simulation
The research introduces a market simulation framework using reinforcement learning agents that can mimic real-world market dynamics and adapt to major market events.
2 shares8 citations todaySource ↗
Curb Appeal with Deep Learning
Incorporating image data into econometric models through deep learning enhances the accuracy of residential real estate price predictions.
4 shares1 citation todaySource ↗
Optimal Rebalancing in AMMs
A new method for optimally rebalancing asset ratios in Dynamic Automated Market Maker pools could potentially increase pool profit and loss by about 25% for a BTC-ETH-DAI pool from July 2022 to June 2023.
4 shares1 citation todaySource ↗
Revisiting String Models of Interest Rates
A revised model of the forward interest rate curve, considering market forces and return correlation, accurately replicates the curve's correlation structure from 1994-2023 with less than 2% error, confirming that perceived time in interest rate markets is a sub-linear function of real time.
3 shares2 citations todaySource ↗
SSRN
Working papers in finance and economics from SSRN.
28 items
Quantitative12
Enhanced Equity Market Strategy
The article introduces a novel stock market strategy that enhances performance by merging a financial stress indicator with sentiment analysis.
29 sharesSource ↗
Extending Financial Portfolio Methods with Deep RL
The paper suggests that deep reinforcement learning can potentially improve traditional portfolio allocation strategies by incorporating contextual data and future rewards.
2 shares1 citation todaySource ↗
Aggregating Autoencoders for Persistent Access Threats
The article presents AEAPT, a deep learning method for detecting and isolating long-term, undetected cyberattacks.
6 sharesSource ↗
Uncertainty in Sentiment Analysis
The study explores the uncertainty in sentiment scores derived from text using advanced language processing models, finding moderate uncertainty in the results.
14 shares1 citation todaySource ↗
Portfolio Replication Insights
The paper introduces a new method for decoding investment portfolio strategies using Dynamic Bayesian Graphical Models, resulting in better portfolio allocation decisions and adaptability to various market conditions.
12 shares3 citations todaySource ↗
Equity Premium Forecasting
Machine learning techniques, while effective in predicting equity premium within sample, struggle to beat the historical average in out-of-sample predictions.
3 sharesSource ↗
China’s Inflation Rate Forecasting
Eight machine learning models, notably the gradient boost decision tree and forecast combination model, excel in predicting China's inflation rate over the autoregressive benchmark.
4 shares1 citation todaySource ↗
Predicting Beta
Machine Learning algorithms enhance the precision of estimating equity betas for private or nontraded assets, particularly for smaller, younger firms with unique capital structures.
2 sharesSource ↗
Deep News Sentiment for Finance
The article discusses the use of neural networks to extract hidden economic factors from large news analytics data, showing superior performance in GDP growth forecasting and asset return analysis.
2 sharesSource ↗
Behavioral Diversification in Portfolios
The article introduces a new simulation of a diversified portfolio based on consumer products, using linear regression and Monte Carlo Simulation, advocating for a consumer-behavior approach in portfolio structuring.
2 sharesSource ↗
Comparative Study of Portfolio Risk Management
The article suggests a new method for portfolio risk management and capital allocation, combining value-at-risk with other statistical measures, proving its effectiveness in reducing potential portfolio losses.
2 sharesSource ↗
Arbitrage Risk in MAX Effect
In the Korean stock market, the MAX effect, or the highest daily return from the previous month, is only significant in overpriced stock groups.
28 sharesSource ↗
Financial16
Forecasting Trading Costs
The research analyzes trading costs, revealing that large, complex trades can be executed affordably and that trade risk value and complexity extend trade horizons.
17 sharesSource ↗
Pricing Liquidity Risk
The research investigates the pricing of liquidity factors in the US stock market, demonstrating that models with a liquidity factor outperform those with a size factor.
5 shares1 citation todaySource ↗
Extreme Liquidity
The article proposes a crypto asset portfolio model that adjusts for liquidity to improve effectiveness and reduce discontinuity.
2 sharesSource ↗
US Treasury Yield Forecast
The study introduces a forecasting model for predicting the 10-Year US Treasury Yield based on variables like exchange rates and crude oil prices.
7 sharesSource ↗
Intellectual Capital Growth Modeling
The paper discusses a new funding mechanism that uses intellectual capital money to stimulate the generation and exploitation of intellectual capital.
3 sharesSource ↗
Market Clearing
The article investigates the role of firms in providing shares to passive investors, particularly in response to index funds' buying.
6 shares2 citations todaySource ↗
Global US Stock Integration
The research finds that US stocks with less global integration can improve portfolio diversification and match international index portfolios in risk-adjusted returns and tail risk.
86 sharesSource ↗
Asymmetric Info in Asset Markets
The study suggests that reducing information asymmetry in secondary asset markets could potentially harm economic welfare.
152 sharesSource ↗
PL Attribution Options
The paper disputes the belief that the gap between implied and realized volatility is the main factor in profit and loss for delta-hedged options, proposing a new formula for understanding this difference.
169 sharesSource ↗
Strategic Mutual Fund Convergence
The research finds a trend towards similar allocation strategies in equity mutual funds globally, especially among funds managed by large financial institutions.
89 sharesSource ↗
Forecasting TSEC Volatility
The study compares GARCH family models and EWMA models to identify the best algorithm for predicting volatility in Taiwan's stock market, using data from 1997 to 2023.
2 sharesSource ↗
Issues with Implied Volatilities
OptionMetrics records stock options prices at 359 p.m., not 400 p.m., causing changes in implied volatility spreads and affecting stock comovement, especially during the COVID-19 pandemic.
2 sharesSource ↗
Improved Volatility Strategy
An enhanced strategy for volatility-managed portfolios, based on Moreira and Muir 2017's formation, results in significant real-time performance improvement, including 148 Sharpe ratio increases and 165 positive abnormal returns.
2 sharesSource ↗
Domain Influence on Diversification
A study of 251 US retail investors found diversification errors in the gain domain but not in the loss domain, supporting a loss-attention hypothesis.
2 sharesSource ↗
False Discoveries in Currency Analysis
A new method, robust to data dependence and estimation errors, is developed to assess predictive models' performance, when applied to currency technical trading rules, it yields a Sharpe ratio around one for about 50 years.
2 sharesSource ↗
Media Sentiment and Asset Allocation
US media sentiment about foreign countries affects domestic investors' international asset allocation, with negative media coverage leading to reduced flows to international mutual funds targeting the country.
2 sharesSource ↗
RePEc
Economics working papers from RePEc's NEP field reports.
4 items
Finance4
Global Impact on Volatility Persistence
Global factors significantly influence the local volatility persistence in equity indices of 17 developed economies.
24 sharesSource ↗
Insider Trading Strategies
Seyhun's 1986 study indicates that insider buying often leads to positive future returns, while insider selling slightly hints at negative returns, possibly due to liquidity needs.
12 sharesSource ↗
Accruals-Cash Flow Evaluation
This research clarifies misconceptions about the role of accruals in informative earnings, introducing a new analysis that recognizes non-cash accruals as parts of earnings that do not involve cash flows.
9 sharesSource ↗
Forecasting CPI
The study enhances the precision and promptness of Consumer Price Index (CPI) forecasts by using a large Chinese news corpus and Internet search data, and combining penalized regression and mixed-frequency data sampling methods.
9 sharesSource ↗
Machine learning
The general machine-learning papers the letter carried in 2023-25.
8 items
Recently Published2
Rashomon Partitions: Estimating Heterogeneity
Estimating Heterogeneity: The study introduces Rashomon Partition Sets, a new method for partitioning covariate space in statistical analyses, which includes all partitions with posterior values near the maximum, allowing for more robust conclusions.
8 shares6 citations todaySource ↗
Gecko: Compact Text Embeddings
Compact Text Embeddings: Gecko is a new text embedding model that improves knowledge extraction from large language models, surpassing other models in the Massive Text Embedding Benchmark.
172 shares90 citations todaySource ↗
Historical Trending6
LightGaussian Compression
LightGaussian is a new method that converts 3D Gaussians into a more compact format, enhancing efficiency in real-time neural rendering and reducing storage needs.
519 shares660 citations todaySource ↗
Longform Factuality
The Search-Augmented Factuality Evaluator (SAFE) method uses large language models to assess the accuracy of long-form factual content, achieving superior rating performance.
323 shares172 citations todaySource ↗
FP Deep Learning Quantization
A study finds that FP8 data formats are superior to INT8 in post-training quantization, offering better workload coverage, model accuracy, and versatility across various network architectures.
100 shares51 citations todaySource ↗
Visual LVLM Grounding
The Rephrase, Augment and Reason (RepARe) framework enhances the performance of large vision-language models in zero-shot tasks by rephrasing questions and extracting image details.
127 shares14 citations todaySource ↗
Unsupervised Diffusion Segmentation
A new method using self-attention layers in stable diffusion models achieves superior zero-shot segmentation without annotations, outperforming previous methods on the COCO-Stuff-27 dataset.
107 shares171 citations todaySource ↗
Riemannian Laplace Approximation
A recent improvement to the Laplace Approximation, which uses a Gaussian distribution to approximate a target density, corrects previous biases and narrow approximations, leading to practical improvements in experiments.
75 shares11 citations todaySource ↗
GitHub
Repositories the letter featured.
8 items
Finance5
Unified Time Series Forecasting Transformers
The piece investigates a combined training approach for universal time series forecasting transformers.
283 shares
PyTorch Implementation for StockFormer
The article showcases a PyTorch implementation of the StockFormer paper, which studies hybrid trading machines using predictive coding.
62 shares
Lightweight LLM Evaluation Suite
The article presents LightEval, a lightweight evaluation suite for LLM, used by Hugging Face along with the new LLM data processing library datatrove and LLM training library nanotron.
267 shares
Python Package for TradingView Screeners
The piece explores a Python package that enables users to develop TradingView screeners.
112 shares
Python Library for Nasdaq Data Links API
The article talks about a Python library that enables access to Nasdaq Data Links' RESTful API.
379 shares
Trending3
Python QuestDB InfluxDB Client
A Python client has been created for QuestDB's InfluxDB Line Protocol.
47 shares
Rust Market Simulation Library
A Rust-based market simulation library with a Python API has been developed.
13 shares
Local and API Model Experiments
Local and API-available models are being used in ongoing experiments.
601 shares
X / Twitter
Posts from quant researchers on X.
12 items
Quantitative6
Turing Institute Report on Language Models in Finance
The report from the Alan Turning Institute explores the application of large language models in finance.
0 shares
Machine Learning for Merger Arbitrage
A recent study utilizes machine learning to enhance the success and potential returns of merger arbitrage trades.
8 shares
Using Option Market Data to Predict ETF Returns
The research paper proposes a novel trading strategy, suggesting that alterations in the implied volatility of ETF options can forecast returns on the underlying ETF.
7 shares
Weekly Recap of New Research in Finance
The latest weekly summary emphasizes new research in fields like asset pricing, machine learning, market microstructure, and options.
6 shares
Machine Learning and Data Science Notes
Ott Toomet from the University of Washington shares extensive lecture notes on machine learning and data science.
5 shares
Equity Risk Premium Update
Aswath Damodaran's 2024 paper update explores the elements affecting the equity risk premium and ways to calculate it.
2 shares
Miscellaneous6
ML DL, AI in Asset Management
The article explores the use of Machine Learning, Deep Learning, and AI in the field of Asset Management.
1 shares
Mathematics of Neural Networks
Bart Smets of Eindhoven University shares lecture notes on the mathematical aspects of Neural Networks for advanced students.
1 shares
Generative AI Language Model Market Map
The article introduces the fast-expanding Generative AI Large Language Model Infrastructure Stack and its market landscape.
0 shares
Info for Traders
The article provides useful information for individuals involved in trading.
0 shares
Linear Algebra Review
The article provides a comprehensive review of the subject of linear algebra.
0 shares