ML-QuantSubscribe

Quant LetterNo. 57

July 2024, Week 3

149 items across 10 sections, as sent to readers on 17 July 2024. Paper titles open their ML-Quant page; ↗ goes to the source.

arXiv

Quantitative-finance and ML-for-finance preprints from arXiv.

8 items

Finance4

Economics3

01

Financial Network Shock Propagation

The study introduces a new way to predict the impact of shocks on a network node, using a vector autoregressive model in the context of the electronic Interbank Deposit Market.

5 sharesSource ↗

02

Seoul Urban Mobility Resilience

The study finds that urban areas with greater economic complexity were more resilient during the 2018 Seoul heat wave, indicating potential population growth in these areas due to global warming.

2 shares5 citations todaySource ↗

03

Elderly Treatment and Cultural Transmission

The study investigates the link between elderly treatment, production, and cultural transmission, indicating a complex relationship between respect for the elderly and various economic factors.

2 sharesSource ↗

Crypto & Blockchain1

01

Stablecoin Transparency and Market Impact

The study shows that in times of market instability, the transparency of the USDC cryptocurrency leads to quick market responses, while the lack of transparency of USDT acts as a buffer against immediate effects. This suggests that investors may seek safety in less transparent cryptocurrencies during turbulent times.

4 shares4 citations todaySource ↗

SSRN

Working papers in finance and economics from SSRN.

27 items

Quantitative19

02

Python Trading Strategies

The author appreciates Kakushadze and Serur for their 151 trading strategies and shares a Python version of these strategies on Github.

8 sharesSource ↗

03

Market Liquidity Determinants

The study applies machine learning to identify factors affecting equity market liquidity, uncovering complex relationships between market liquidity and placement characteristics.

3 sharesSource ↗

11

APT and Arbitrage

The Arbitrage Pricing Theory (APT) enables potential arbitrage opportunities due to the absence of a positive pricing function requirement.

3 sharesSource ↗

12

PE Optimism Q1’24

Confidence among private equity (PE) firms has significantly increased over the past two years, with Q1 2024 showing renewed optimism despite market fluctuations.

3 sharesSource ↗

13

Multi-Agent Stock Simulation

A new multi-agent distributed stock exchange simulation environment (DSXE) has been developed to model global financial markets, enabling large-scale simulations and successful fragmented market modeling.

2 sharesSource ↗

14

Healthcare Fraud Detection

Machine learning techniques are being used to detect healthcare fraud in the US, with a study analyzing over 558,211 records using various ML models.

2 sharesSource ↗

15

AI Model Explainability

The article discusses the legal challenges posed by the use of complex AI and machine learning models in decision-making processes due to their lack of traceability.

2 sharesSource ↗

16

Rethinking Forex

The author criticizes the current financial system for its inefficiency and frequent crises, suggesting the need for a new, more efficient system.

2 sharesSource ↗

17

Humanising Portfolio Selection

The paper proposes a unified approach to active portfolio selection, demonstrating how investor subjectivity can improve portfolio performance.

2 sharesSource ↗

18

Insider Ownership in Japan

A study reveals that insider ownership significantly increases default risk in Japanese firms, based on data from 2004-2019.

2 sharesSource ↗

Financial8

03

Geometric Insights on Portfolio Construction

The article argues that the equally weighted portfolio is usually less preferable than the mean-variance portfolio, based on the influence of the covariance matrix's condition number on the αweight angle in portfolio optimizations.

3 sharesSource ↗

04

Semivolatility portfolios

The article proposes the use of semi-volatility-managed portfolios to enhance the performance of momentum portfolios by controlling skewness and downside volatility.

2 sharesSource ↗

05

Dual Dominance

The article explores the relationship between the Markowitz mean-variance model and the Ziemba capital growth model, offering insights into model-based portfolio construction.

2 sharesSource ↗

06

CNN2D-MV Stock Model

The article introduces a hybrid model that combines a 2D convolutional neural network with the mean-variance model to improve asset selection and portfolio performance.

2 sharesSource ↗

07

Smarter Beta Investing

The article suggests integrating sustainable factors into traditional investing methods without affecting financial performance or diversification, and offers ways to correct sustainable bias in traditional long-short MSCI style factor portfolios.

2 sharesSource ↗

08

Debt and Trade: Global View

Global View: The study reveals that multinational corporations often raise debt capital outside their home country to hedge against exchange rate fluctuations and align with their supply chain markets, in addition to accessing deeper financial markets.

2 sharesSource ↗

RePEc

Economics working papers from RePEc's NEP field reports.

30 items

Finance10

01

Forecasting Accuracy in Markets

The study examines the fluctuation in commodity returns and financial market index using different models, finding that the effectiveness of models varies with assets like gold, cocoa, and the S&P500 Index.

31 sharesSource ↗

02

Cryptocurrency Volatility vs. Asset Classes

The paper explores factors affecting future volatility in the cryptocurrency ecosystem, discovering that positive market returns increase price volatility, contrary to traditional financial studies.

25 sharesSource ↗

03

Asian Stock Exchange Volatility with Bitcoin and Oil

The research looks into the volatility of the Asian stock market in relation to Bitcoin and global crude oil prices, showing volatility clustering and varying volatility spillover from crude oil and Bitcoin to different Asian stock exchanges.

24 sharesSource ↗

04

Market Conditions and Liquidity Shocks

The study investigates the impact of market conditions, volatility, and liquidity shocks on arbitrage profits during pre-COVID and COVID periods, concluding that high volatility and low liquidity during COVID made arbitrage unprofitable.

23 sharesSource ↗

05

Performance of Multi-Asset Funds During Crises

The paper studies the performance of multi-asset funds investing internationally, finding that these funds underperformed from 2004 to 2021, but performed better during market crises, with bond-focused funds doing better in non-crisis periods and equity-focused funds doing better during crises.

22 sharesSource ↗

Statistical6

01

XGBoost for LGD Approximation

The study investigates ways to improve the estimation of Loss Given Default using machine learning and European mortgage data, particularly when cash-flow data is scarce.

25 sharesSource ↗

02

Markov-Switching Trees

The research suggests a method that merges decision trees and time series modeling, using eight seasons of NFL data to predict play calls, bridging the gap between machine learning and statistics.

23 sharesSource ↗

03

Multi-Level Retail Loan Optimization

The article explores a multi-level portfolio selection model for retail-banking loans, aiming to optimize risk and return at both loan and bank levels, and compares the optimized portfolio with the original for potential benefits.

20 sharesSource ↗

04

Conditional Feature Importance

The research proposes a method that merges the conditional predictive impact framework with sequential knockoff sampling in machine learning, emphasizing the need to consider variable importance before and after adjusting for covariates.

15 sharesSource ↗

05

Machine Learning for NPLs Profitability

The article presents a machine learning approach to predict the recovery rate of non-performing loans, aiming to minimize the lemon discount caused by information asymmetry between banks and investors.

14 sharesSource ↗

06

AI Readiness in Economies

The study uses machine learning to identify factors affecting AI readiness in businesses in 40 countries, highlighting scientific research output, internet infrastructure, and public consumption expense as common influential factors.

13 sharesSource ↗

Machine Learning3

01

Stock Index Prediction

The research uses sentiment analysis and machine learning to determine that investor sentiment and exchange rates greatly impact the Shanghai Composite Index.

14 sharesSource ↗

02

Efficient Hyperparameter Tuning

The study presents the sequential random search (SQRS) method for more efficient hyperparameter tuning in machine learning by discarding poor parameter configurations early.

13 sharesSource ↗

03

Forecasting Standard Dominance

The paper applies machine learning to predict the results of standard battles in the Chinese solid-state lighting industry, concluding that strong alliances, patent application experience, and marketization level increase a firm's likelihood of success.

12 sharesSource ↗

Deep Learning1

01

FinBERT and LSTM for Stock Price Prediction

The article discusses a hybrid model that combines BERT and LSTM for predicting stock prices. This model surpasses traditional methods by including financial news sentiment analysis and technical indicators, allowing for accurate predictions of significant stock price fluctuations.

12 sharesSource ↗

Historical Trending10

02

Investor Attention and Volatility

The article investigates the volatility puzzle in China's stock market, attributing it to individual investor attention and the role of securities analysts in reducing information asymmetry.

16 sharesSource ↗

05

Transnational Bid-Rigging Detection

Machine learning and statistical screening can accurately detect bid-rigging cartels, but their effectiveness decreases when used on data from different countries due to institutional variations.

24 sharesSource ↗

06

Anomalies and Return Predictability

The study shows that portfolio returns based on long-short anomaly can predict overall market returns, due to asymmetric limits of arbitrage and overpricing correction persistence.

116 sharesSource ↗

07

Factor Timing in China

The paper proposes a deep learning strategy using 146 factors, which is effective and robust in the unique structure of the Chinese stock market.

52 sharesSource ↗

08

Forecasting Chinese Macroeconomy

The research finds that mixed-frequency factor models are better at forecasting the Chinese macroeconomy than traditional models, except during the Global Financial Crisis.

10 sharesSource ↗

09

Currency Risk Pricing

The study finds that 85% of risk in individual currencies does not affect their risk premiums when using interest differential, trend, and mean reversion signals to construct a portfolio.

21 sharesSource ↗

10

War in Ukraine Analysis

The research examines the topics and sentiments of Ukrainian Telegram users during the early stages of the war in Ukraine, emphasizing the importance of social media analytics.

9 sharesSource ↗

Machine learning

The general machine-learning papers the letter carried in 2023-25.

20 items

Recently Published10

01

Optimizing NLP Systems

The article explores a method to enhance Natural Language Processing systems by simultaneously optimizing language model weights and prompting strategies, leading to significant improvements in tasks like multi-hop QA and mathematical reasoning.

60 shares59 citations todaySource ↗

02

Human-like Memory

The study presents EM-LLM, a new approach that incorporates elements of human episodic memory into Large Language Models, enabling them to manage infinite context lengths efficiently and outperform existing models in tasks like PassageRetrieval.

28 shares69 citations todaySource ↗

03

Transformers for Chatbots

The research builds a Transformer that replicates the ELIZA program, a traditional rule-based chatbot, to gain insights into the preferred mechanisms of Transformer-based chatbots.

19 shares2 citations todaySource ↗

04

Anomaly Detection with LLMs

The paper introduces a two-stage reasoning framework for identifying and mitigating out-of-distribution failure modes in robotic systems using large language models, comprising a quick binary anomaly classifier and a slower fallback selection stage.

17 shares105 citations todaySource ↗

05

Topological Bounds for Algorithms

The research proposes a new set of topology-based complexity notions that correlate with the generalization gap in deep neural networks, offering a computationally efficient way to predict generalization without test data, and surpassing existing topological bounds across various datasets and models.

15 shares12 citations todaySource ↗

06

Spider2-V: Data Science Automation

Data Science Automation: Spider2-V is a benchmark introduced to assess the performance of multimodal agents in automating data science and engineering workflows, showing that current models have difficulties in fully automating these workflows.

11 shares61 citations todaySource ↗

08

LNS Benchmarking for Multi-Agent Path Finding

The research provides a balanced comparison of Multi-Agent Path Finding solvers and presents a new strategy for neighborhood selection that enhances runtime efficiency in large maps with numerous agents.

10 shares3 citations todaySource ↗

09

MIA Bird's Eye View Mapping

The paper presents Map It Anywhere, a data engine that uses crowd-sourced mapping platforms to create a dataset for Bird's Eye View map prediction, increasing zero-shot performance by 35%.

9 shares6 citations todaySource ↗

10

Transformer Circuit Metrics

The authors explore the difficulties in evaluating the performance of neural network 'circuits', emphasizing the sensitivity of current methods to changes in the ablation methodology and the need for clearer claims about circuits.

9 shares19 citations todaySource ↗

Historical Trending10

02

Exploration-Based Optimization for LLM Agents

The study explores an exploration-based trajectory optimization (ETO) method to enhance the performance of Large Language Models (LLMs) by learning from their exploration mistakes.

245 shares206 citations todaySource ↗

03

SVGDriven Image Understanding with LLMs

The research examines the capability of Large Language Models (LLMs) to interpret images by transforming them into Scalable Vector Graphics (SVG) and assessing the LLMs on various computer vision tasks.

65 shares8 citations todaySource ↗

04

Lie Group Decompositions for Equivariant NNs

The paper introduces a framework for dealing with Lie groups and their homogeneous spaces, showing how to parametrize convolution kernels to create models that are equivariant to affine transformations.

60 shares15 citations todaySource ↗

05

Building RAG Chatbots

The article introduces the FACTS framework for developing Retrieval Augmented Generation (RAG)-based chatbots, and presents empirical results on the balance between accuracy and latency in large and small LLMs.

51 shares37 citations todaySource ↗

08

Speaker Diarization with LLMs

DiarizationLM, a new framework, improves transcript readability and reduces word diarization error rate by using large language models to post-process speaker diarization system outputs.

27 shares34 citations todaySource ↗

09

Sparse-View 3D Gaussian Splatting

A new perspective for 3D Gaussian Splatting improves sparse-view 3DGS by identifying and suppressing inaccurate reconstruction, achieving high-quality novel view synthesis.

27 shares176 citations todaySource ↗

10

Cultural Competence in Text-to-Image Models

A new framework assesses the cultural competence of Text-to-Image models, revealing significant gaps in cultural awareness and providing insights into the cultural diversity of model outputs.

23 shares45 citations todaySource ↗

Papers with code

Papers that shipped their code, from the Papers with Code feed (2023-25).

9 items

Trending4

01

Voice Interaction Models

The report presents FunAudioLLM, a new model aimed at enhancing voice communication between humans and large language models.

2,037 shares

02

OnDevice Language Models

The MobileLLMLS models demonstrate an accuracy improvement of 0.70.8 over the MobileLLM 125M350M models.

648 shares

03

Multimodal Story Generation

A new multimodal attention sink mechanism is suggested for efficiently creating stories with up to 25 sequences, using only 10 for training.

142 shares

Rising5

03

QGaLore: Quantized GaLore with INT4 Projection

Quantized GaLore with INT4 Projection: QGalore enhances memory efficiency by merging quantization and low-rank projection, offering improvements over the GaLore system.

98 shares

05

MAVIS: Mathematical Visual Instruction Tuning

Mathematical Visual Instruction Tuning: Enhancements in visual encoding of math diagrams, alignment of diagrams and language, and mathematical reasoning abilities are required in MLLMs.

48 shares

GitHub

Repositories the letter featured.

10 items

Finance5

01

HighFrequency Trading with Data Science

The article explores the application of data science and machine learning in creating high-frequency trading strategies with full orderbook tick data.

1,867 shares

02

Online Machine Learning Python

The piece provides a tutorial on implementing online machine learning using the Python programming language.

4,884 shares

03

Investment Research

The article discusses how to make investment research available to everyone, irrespective of their geographical location.

26,626 shares

04

Rust Machine Learning

The article introduces a new machine learning framework developed using the Rust programming language.

3,552 shares

05

HighPerformance Trading

The piece showcases a high-performance trading library, developed in Mojo and C, designed to simplify quantitative trading.

33 shares

Trending5

05

Tegon: Open-Source AI for Jira Linear

Open-Source AI for Jira Linear: Tegon is a new open-source, AI-powered project management tool, offering an alternative to Jira and Linear.

1,101 shares

News

Industry news: funds, hiring, markets and regulation.

20 items

Quantitative10

01

AI Fund Launch

Astant Global Management, a London-based investment manager, is set to launch a new AI-based quantitative hedge fund strategy later this year.

6 shares

02

Archegos Founder Convicted

Archegos Capital Management's founder, Bill Hwang, has been convicted of securities fraud and market manipulation, leading to massive losses for global investment banks.

5 shares

03

TNS Includes LTSE Feeds

Transaction Network Services will now support the new range of MEMOIR market data feeds from the LongTerm Stock Exchange.

4 shares

04

Statar Capital Reduces Losses

Ron Ozer, a natural gas trading specialist, has managed to cut losses at his Miami-based hedge fund, Statar Capital, following significant declines earlier in Q2.

4 shares

05

Quant Funds: FinTech

FinTech: Don Silva's article on Medium explores the convergence of finance and technology in Quant Funds.

4 shares

07

Longshort Equity Q2 Outperforms

Equity long-short strategies had an average increase of 2.7% in Q2, outperforming other strategies, as per Unlimited's Hedge Fund Barometer.

4 shares

08

South Korea Hedge Fund Payment

South Korea is contesting a Hague court's decision that it must pay $32m to US hedge fund Mason Capital Management for interfering in a 2015 Samsung merger.

3 shares

09

Segantii Returns Client Capital

Segantii Capital Management has reportedly returned over 90% of client capital in less than two months after deciding to refund.

3 shares

Miscellaneous10

02

Liquidnet appoints Head of International

Liquidnet has hired Jeffrey Crane as Head of International in the Americas, reporting to Alan Polo, Head of Equity Sales and Trading Americas.

3 shares

04

Hedge fund software bets hit record low

Global hedge funds have cut their exposure to US software stocks to multi-year lows following a tech sector selloff, says Morgan Stanley's prime brokerage division.

2 shares

05

China tightens shortselling restrictions

China's securities regulator has imposed more restrictions on short-selling and pledged stricter oversight of computer-driven trading to bolster the struggling stock market.

2 shares

06

Digital Assets Funds Inflow

CoinShares' Digital Assets Fund Flows Weekly Report shows that digital asset investment products had the fifth highest weekly total on record, with inflows of 1.44bn.

2 shares

07

SkyBridge Client Exit Limits

SkyBridge Capital, owned by Anthony Scaramucci, has restricted client withdrawals from its crypto-focused hedge fund despite high returns, as per a Bloomberg report.

2 shares

08

DTCC FICC VaR Calculator

The Depository Trust & Clearing Corporation has launched a public Value at Risk calculator to assist market participants in evaluating potential margin and clearing fund obligations.

2 shares

10

GoldenTree €404m CLO Closure

GoldenTree Asset Management has completed a €404m collateralised loan obligation managed by GLM III, marking the 28th CLO issued under the firm's GLM CLO strategy, totalling over 15bn.

1 shares

Podcasts

Episodes on markets, quant methods and economics.

10 items

Quantitative5

01

Market Rebellion with Jon Najarian

Ex-Chicago Bears player Jon Najarian talks about his shift from sports to trading, the difficulties in the cannabis industry, and the need for disciplined trading and diversification in an unstable market.

19 shares

02

EM Fixed Income vs US elections

EM strategists Jonny Goulden and Saad Siddiqui discuss the effect of the US elections on EM markets and supportive global data, strictly prohibiting the use of J.P. Morgan Data in third-party AI systems.

8 shares

03

EM Fixed Income Focus vs US elections

Jonny Goulden and Saad Siddiqui update on EM markets, highlighting the impact of US elections and global data, and stress on the privacy of J.P. Morgan Data in AI systems.

8 shares

04

Michael Mauboussin Market Concentration

Michael Mauboussin, Head of Consilient Research at Counterpoint Global, talks about decision-making, behavioral economics, and investing, covering topics like public vs. private equity, luck vs. skill, and tips for budding investment professionals.

8 shares

05

Global FX On Japan intervention

Global FX Strategists discuss the future of yen flows, the dollar's response to CPI and payrolls, and the implications of the latest data for EUR, GBP, and Scandi FX, strictly prohibiting the use of J.P. Morgan Data in third-party AI systems.

7 shares

Related5

01

Investing in Cannabis

Aaron Edelheit, CEO of Mindset Capital, discusses the potential for cannabis legalization at the federal level, the industry's challenges and opportunities, and his interest in low dose hemp beverages.

7 shares

02

Global Market Predictions

Economist David Rosenberg expresses optimism for markets outside the S&P 500, pointing out opportunities in Japan, India, and commodities, while cautioning about a possible recession.

6 shares

03

Unified Trading Framework

Lorenzo Ravagli of JP Morgan suggests a new approach for trading the volatility skew premium.

4 shares

04

Oil Balance Dichotomy

Natasha Kaneva, Head of Global Commodities Research, forecasts a 1.0 mbd oil liquids deficit in 3Q, maintains her prediction of Brent oil reaching $90 by September, and expects a drop to mid-$60s in 4Q25.

2 shares

05

Tian Yang's Trading Insights

MacroVoices hosts Erik Townsend and Patrick Ceresna invite Variant Perception CEO Tian Yang to discuss leading indicators and the most advantageous trades currently.

2 shares

Blogs

Posts from quant and economics blogs and newsletters.

4 items

Quantitative2

01

Exotic Markets

Russell Korgaonkar discusses Man AHL's investment in unusual markets, highlighting potential benefits for investors and predicting future trends.

2 shares

02

TSV Strategy

The article explores the use of Time in simplifying the analysis of price fluctuations and spotting trading opportunities.

1 shares

Related2

02

Independent Event Probability

The article delves into the idea of events that occur independently of human beliefs or perceptions.

0 shares

X / Twitter

Posts from quant researchers on X.

11 items

Quantitative5

01

Practical Uses of Derivatives

Fabozzi's research illustrates the practical applications of derivatives in areas like asset allocation and liquidity management.

7 shares

03

Statistical Arbitrage with Oil Futures

Fanelli's research presents a statistical arbitrage portfolio involving different oil futures, demonstrating significant performance after costs.

5 shares

04

Quantile Regression for Factor Modeling

A study on Quantile Regression and Equity Factor Modeling explores the application of the 3 and 5 Factor model in Taiwan's stock market.

3 shares

05

Liquid Alternative Strategies

ManGroup's report suggests that liquid alternative strategies, such as trend-following and long-short quality stocks, could potentially replace bonds.

1 shares

Miscellaneous6

02

Firm Profitability Decomposition

The article reviews a study by Han et al., suggesting that investors frequently undervalue the impact of firm-specific factors on profitability, leading to substantial returns.

0 shares

04

Bayesian Sports Models in R

The author has released a new book titled Bayesian Sports Models in R, featuring colorized R and Stan code and a complete R code download package.

0 shares

    Type to search. Try rough volatility, LLM agents or FinGPT.

    ↑↓ move↵ openesc closeFull search page