Machine Learning in Finance Workshop 2020
The workshop is organized by:
On Friday Aug 7th, materials including research abstracts, presentations and videos will be released to all registered attendees.
Please RSVP (details below) to receive more announcements about the virtual offerings.
Participating Researchers/Practitioners
Deep Learning Under Distribution Shift
Abstract:
We might hope that when faced with unexpected inputs, well-designed software systems would fire off warnings. However, ML systems, which depend strongly on properties of their inputs (e.g. the i.i.d. assumption), tend to fail silently. Faced with distribution shift, we wish (i) to detect and (ii) to quantify the shift, and (iii) to correct our classifiers on the fly — when possible. This talk will describe a line of recent work on tackling distribution shift. First, I will focus on recent work on label shift, a more classic problem, where strong assumptions enable principled methods. Then I will discuss how recent tools from generative adversarial networks have been appropriated (and misappropriated) to tackle dataset shift—characterizing and (partially) repairing a foundational flaw in the method.
Bio:
Zachary Chase Lipton is the BP Junior Chair Assistant Professor of Operations Research and Machine Learning at Carnegie Mellon University. His research spans core machine learning methods and their social impact and addresses diverse application areas, including clinical medicine and natural language processing. Current research focuses include robustness under distribution shift, breast cancer screening, the effective and equitable allocation of organs, and the intersection of causal thinking and the messy high-dimensional data that characterizes modern deep learning applications. He is the founder of the Approximately Correct blog (approximatelycorrect.com) and a co-author of Dive Into Deep Learning, an interactive open-source book drafted entirely through Jupyter notebooks. Find on Twitter (@zacharylipton) or GitHub (@zackchase).
Forecasting Firm Material Event Sequences from SEC 8-K Reports
Abstract:
Based on a unique sequence-to-sequence formulation, we propose a hybrid Transformer model to forecast firm material event sequences, by utilizing the firm's SEC 8-K current reports and financial ratios. Our proposed model demonstrates superior prediction performance compared against traditional sequence-to-sequence models and task-specific Markov Chain Monte Carlo simulations.
Bio:
Zhu (Drew) Zhang is an associate professor of information systems and business analytics at Ivy College of Business, Iowa State University. He earned his Ph.D. in computer science from University of Michigan. With his primary expertise in natural language processing and machine learning, his research program in AI for Finance focuses on harnessing the power of unstructured data such as text.
Agency MBS Prepayment Model Using Neural Networks
Abstract:
Artificial intelligence can reduce model fitting times from months to hours, significantly improving modeling efficiency and enabling true model back-testing and timelier understanding of prepayment trends. Our AI prepayment model has demonstrated higher model accuracy and agility than our traditional model. It overcame high-dimensionality and high-nonlinearity issues associated with prepayment modeling. The AI prepayment model was able to detect new and often subtle prepayment signals that eluded traditional modeling approaches.
Bio:
David Zhang is a Managing Director and Head of Securitized Products Research at MSCI. His team is responsible for developing models and analytics to support investment analysis, risk management, and regulatory compliance. Since joining MSCI in December 2016, his group has developed new models for interest rate and mortgage rate, agency MBS prepayment, mortgage credit, ABS, CLO and Chinese ABS. They have also pioneered and published several big data and artificial intelligence researches. Before joining MSCI, Dr Zhang was Managing Director and head of Securitized Products modeling at Credit Suisse for twelve years, responsible for supporting risk, regulatory and client analytics as well as sales/trading quantitative strategies. Dr Zhang’s group developed one of the most widely used MBS models by fixed income institutional investors. Their work was consistently awarded top ranking by various industry and client surveys, including Institutional Investor All-America Research Team ranking in Agency prepayment.
David is the president and board director of TCFA (The Chinese Finance Association). He also serves on board of GCREC (Global Chinese Real Estate Congress), and NY PRMIA (Professional Risk Management International Association). He co-chairs the China Market Committee of SFA (Structured Finance Industry Group). He has published widely in academic and industry journals. Dr Zhang has a Ph.D. from Princeton University.
Machine Learning Applications in Asset Management
Abstract:
The multi-step processes include applying machine-learning techniques to construct portfolio asset allocations by optimizing certain variables including risk, return, duration, other for clusters of investors. Then, applying machine learning techniques to identify key features that influence security selection, and based on the key features, construct and optimize the portfolio holdings. Finally, use NLP and sentiment analysis techniques to asses the impact of real-time events on portfolio holdings and rebalancing.
Bio:
Tugce Karatas is a PhD student in Industrial Engineering & Operations Research Department at Columbia University. She received her B.S. and M.S. degrees from Industrial Engineering Department at Bogaziçi University, Turkey. She is currently working with Prof. Ali Hirsa on machine learning applications in finance. Her research focuses on the application of neural networks and machine learning techniques in a broad range of financial instruments specifically asset managemnet.
Satyan Malhotra is a Managing Partner with Flexstone Partners, which has $7.3billion in assets under management and advisory. He works on the firm’s day-to-day operations, business development and new product offerings. He serves on the US Audit Committee, the Global Advisory Investment Committee, and the US Investment Committee, and the Compensation Committee.
Satyan is a founding Partner of Caspian Private Equity, a predecessor to Flexstone Partners. Prior and concurrent to that, until January 2012, he served as President of Caspian Capital Management (“CCM”) that managed as much as $2.2Bn of alternative assets including fixed income portfolios and quantitative trading strategies. CCM also provided non-discretionary consulting on as much as $3Bn in structured products on funds of hedge fund programs. Prior to joining CCM in 2002, Satyan was a Senior Manager with the Global Risk Management Solutions consulting practice at PricewaterhouseCoopers. In that capacity, he worked with major financial institutions, corporations and treasuries on a variety of risk management, portfolio and strategic initiatives.
Satyan holds a CIBE from Columbia Business School, an MBA in Finance from Virginia Tech, a BA in Economics (Honors) from the University of Delhi and a Financial Risk Manager (FRM) certification from the Global Association of Risk Professionals (GARP).
He serves on various company boards and has authored several research pieces on investments.
Optimal, Truthful, and Private Securities Lending
Abstract:
We consider a fundamental dynamic allocation problem motivated by the problem of securities lending in financial markets, the mechanism underlying the short selling of stocks. A lender would like to distribute a finite number of identical copies of some scarce resource to n clients, each of whom has a private demand that is unknown to the lender. The lender would like to maximize the usage of the resource — avoiding allocating more to a client than her true demand — but is constrained to sell the resource at a pre-specified price per unit, and thus cannot use prices to incentivize truthful reporting. We first show that the Bayesian optimal algorithm for the one-shot problem — which maximizes the resource’s expected usage according to the posterior expectation of demand, given reports — actually incentivizes truthful reporting as a dominant strategy. Because true demands in the securities lending problem are often sensitive information that the client would like to hide from competitors, we then consider the problem under the additional desideratum of (joint) differential privacy. We give an algorithm, based on simple dynamics for computing market equilibria, that is simultaneously private, approximately optimal, and approximately dominant-strategy truthful. Finally, we leverage this private algorithm to construct an approximately truthful, optimal mechanism for the extensive form multi-round auction where the lender does not have access to the true joint distributions between clients’ requests and demands.
Bio:
Emily Diana is a PhD student in Statistics at the Wharton School, University of Pennsylvania, where she is working in the research group of Michael Kearns and Aaron Roth. She holds a B.A. in Applied Mathematics from Yale and an M.S. in Statistics from Stanford. Before graduate school, she worked for two years as a scientific software developer at Lawrence Livermore National Laboratory, focusing on improving the performance of government finite element physics simulation codes. Her current research interests include computational statistics, economic game theory, and differential privacy.
Designing Equitable Risk Models for Lending and Beyond
Abstract:
From credit scoring to evaluating job candidates, statistical models are increasingly used to guide high-stakes human experts, including lenders, managers, judges, and doctors. Researchers and policymakers, however, have raised concerns that these machine-learned algorithms might inadvertently exacerbate societal biases. For example, poorly constructed credit scores might disproportionately harm minority borrowers. To measure and mitigate such potential bias, there's recently been an explosion of competing mathematical definitions of what it means for a risk model to be fair. But there’s a problem: nearly all the prominent definitions of fairness suffer from subtle shortcomings that can lead to serious adverse consequences when used as a design principle. I'll illustrate these problems that lie at the foundation of this nascent field of algorithmic fairness, drawing on ideas from machine learning, economics, and legal theory. In doing so, I hope to offer practitioners a way to make more equitable decisions.
Bio:
Sharad Goel is an Assistant Professor at Stanford University in the Department of Management Science & Engineering, with courtesy appointments in Computer Science, Sociology, and the Law School. He's the founder and director of the Stanford Computational Policy Lab, a group that develops technology to tackle pressing issues in criminal justice, education, voting rights, and beyond. In his research, Sharad looks at public policy through the lens of computer science, bringing a new, computational perspective to a diverse range of contemporary social issues, including policing practices, electoral integrity, online privacy, and media bias. Before joining the Stanford faculty, Sharad completed a Ph.D. in Applied Mathematics at Cornell University, and worked as a Senior Researcher at Microsoft.
Applications of Deep Generative Modeling in Finance
Abstract:
Generative modeling techniques have experienced a resurgence in the machine learning research community over the last five to ten years with Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Normalizing Flows as only a few examples. While publications have largely focused on showing progress in the domains of images and natural language processing, the underlying innovations have implications outside of those narrow applications. In this talk, we will share concrete applications of these techniques that resolve real-world challenges in financial modeling that arise from the difficulties inherent in financial datasets.
Bio:
Achintya Gopal is a Machine Learning Quant Researcher at Bloomberg, where he was previously a Software Engineer. He earned his Masters and Bachelor's degrees in Computer Science at Johns Hopkins University. His research interests are focused on machine learning and financial modeling, particularly the applications of normalizing flows.
Modelling Asset Returns with Quant GANs
Abstract:
Modelling financial time series by using stochastic processes is a challenging task and a central area of research in mathematical finance. In this talk, we will explore how Quant GANs, a data-driven generative model based on generative adversarial networks (GANs), can be used as an alternative. The proposed Quant GAN consist out of a discriminator-generator pair that utilises temporal convolutional networks (TCNs). This architecture choice offers several advantages such as parallelism, stable gradients, guaranteed stationarity of the generated paths and enables the Quant GAN to capture volatility clusters and leverage effects. A numerical study for the S&P 500 index is presented, highlighting that various distributional and dependence properties — for small and large lags — are in excellent agreement.
Bio:
Magnus Wiese is a mathematics Ph.D. student at the University of Kaiserslautern. His research focuses on the application of neural networks in mathematical finance for calibration and generative modelling. Prior to his Ph.D. studies, he interned at J.P. Morgan where he worked on modelling and expanding limited real-world datasets such as equity options by using adversarial training techniques.