Artificial Intelligence Research

@inproceedings{Byrd2019FundAI,

title={Fund Asset Inference Using Machine Learning Methods: What’s in That Portfolio?},

author={David Byrd and Sourabh Bajaj and Tucker Hybinette Balch},

booktitle={The Journal of Financial Data Science},

year={2019}

}

Given only the historic net asset value of a large-cap mutual fund, which members of some universe of stocks are held by the fund? Discovering an exact solution is combinatorially intractable because there are, for example, C(500,30) or 1.4 × 1048 possible portfolios of 30 stocks drawn from the S&P 500. The authors extend an existing linear clones approach and introduce a new sequential oscillating selection method to produce a computationally efficient inference. Such techniques could inform efforts to detect fund window dressing of disclosure statements or to adjust market positions in advance of major fund disclosure dates. The authors test the approach by tasking the algorithm with inferring the constituents of exchange traded funds for which the components can be later examined. Depending on the details of the specific problem, the algorithm runs on consumer hardware in 8 to 15 seconds and identifies target portfolio constituents with an accuracy of 88.2% to 98.6%.

@misc{https://doi.org/10.48550/arxiv.1907.09567,

doi = {10.48550/ARXIV.1907.09567},

url = {https://arxiv.org/abs/1907.09567},

author = {Cohen, Naftali and Balch, Tucker and Veloso, Manuela},

keywords = {Computer Vision and Pattern Recognition (cs.CV), Human-Computer Interaction (cs.HC), Computational Finance (q-fin.CP), Trading and Market Microstructure (q-fin.TR), FOS: Computer and information sciences, FOS: Computer and information sciences, FOS: Economics and business, FOS: Economics and business},

title = {The Effect of Visual Design in Image Classification},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

Financial companies continuously analyze the state of the markets to rethink and adjust their investment strategies (e.g., [8, 4, 15]). While the analysis is done on the digital form of data, decisions are often made based on graphical representations in white papers or presentation slides. In this study, we examine whether binary decisions are better to be decided based on the numeric or the visual representation of the same data. Using two data sets, a matrix of numerical data with spatial dependencies and financial data describing the state of the S&P index, we compare the results of supervised classification based on the original numerical representation and the visual transformation of the same data. We show that, for these data sets, the visual transformation results in higher predictability skill compared to the original form of the data. We suggest thinking of the visual representation of numeric data, effectively, as a combination of dimensional reduction and feature engineering techniques (e.g., [5]). In particular, if the visual layout encapsulates the full complexity of the data. In this view, thoughtful visual design can guard against overfitting, or introduce new features – all of which benefit the learning process, and effectively lead to better recognition of meaningful patterns

@misc{https://doi.org/10.48550/arxiv.1910.12175,

doi = {10.48550/ARXIV.1910.12175},

url = {https://arxiv.org/abs/1910.12175},

author = {Chan, T-H. Hubert and Liang, Zhibin and Polychroniadou, Antigoni and Shi, Elaine},

keywords = {Information Theory (cs.IT), FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Small Memory Robust Simulation of Client-Server Interactive Protocols over Oblivious Noisy Channels},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

We revisit the problem of low-memory robust simulation of interactive protocols over noisy channels. Haeupler [FOCS 2014] considered robust simulation of two-party interactive protocols over oblivious, as well as adaptive, noisy channels. Since the simulation does not need to have fixed communication pattern, the achieved communication rates can circumvent the lower bound proved by Kol and Raz [STOC 2013]. However, a drawback of this approach is that each party needs to remember the whole history of the simulated transcript. In a subsequent manuscript, Haeupler and Resch considered low-memory simulation. The idea was to view the original protocol as a computational DAG and only the identities of the nodes are saved (as opposed to the whole transcript history) for backtracking to reduce memory usage. In this paper, we consider low-memory robust simulation of more general client-server interactive protocols, in which a leader communicates with other members/servers, who do not communicate among themselves; this setting can be applied to information-theoretic multi-server Private Information Retrieval (PIR) schemes. We propose an information-theoretic technique that converts any correct PIR protocol that assumes reliable channels, into a protocol which is both correct and private in the presence of a noisy channel while keeping the space complexity to a minimum. Despite the huge attention that PIR protocols have received in the literature, the existing works assume that the parties communicate using noiseless channels. Moreover, we observe that the approach of Haeupler and Resch to just save the nodes in the aforementioned DAG without taking the transcript history into account will lead to a correctness issue even for oblivious corruptions. We resolve this issue by saving hashes of prefixes of past transcripts. Departing from the DAG representation also allows us to accommodate scenarios where a party can simulate its part of the protocol without any extra knowledge (such as the DAG representation of the whole protocol). In the the two-party setting, our simulation has the same dependence on the error rate as in the work of Haeupler, and in the client-server setting it also depends on the number of servers. Furthermore, since our approach does not remember the complete transcript history, our current technique can defend only against oblivious corruptions.

@misc{https://doi.org/10.48550/arxiv.1906.02312,

doi = {10.48550/ARXIV.1906.02312},

url = {https://arxiv.org/abs/1906.02312},

author = {Vyetrenko, Svitlana and Xu, Shaojie},

keywords = {Trading and Market Microstructure (q-fin.TR), Machine Learning (cs.LG), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Risk-Sensitive Compact Decision Trees for Autonomous Execution in Presence of Simulated Market Response},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

We demonstrate an application of risk-sensitive reinforcement learning to optimizing execution in limit order book markets. We represent taking order execution decisions based on limit order book knowledge by a Markov Decision Process; and train a trading agent in a market simulator, which emulates multi-agent interaction by synthesizing market response to our agent’s execution decisions from historical data. Due to market impact, executing high volume orders can incur significant cost. We learn trading signals from market microstructure in presence of simulated market response and derive explainable decision-tree-based execution policies using risk-sensitive Q-learning to minimize execution cost subject to constraints on cost variance.

@article{https://doi.org/10.48550/arxiv.1906.12010,

doi = {10.48550/ARXIV.1906.12010},

url = {https://arxiv.org/abs/1906.12010},

author = {Balch, Tucker Hybinette and Mahfouz, Mahmoud and Lockhart, Joshua and Hybinette, Maria and Byrd, David},

keywords = {Trading and Market Microstructure (q-fin.TR), Computer Science and Game Theory (cs.GT), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {How to Evaluate Trading Strategies: Single Agent Market Replay or Multiple Agent Interactive Simulation?},

publisher = {arXiv},

year = {2019}

copyright = {arXiv.org perpetual, non-exclusive license}

}

We show how a multi-agent simulator can support two important but distinct methods for assessing a trading strategy: Market Replay and Interactive Agent-Based Simulation (IABS). Our solution is important because each method offers strengths and weaknesses that expose or conceal flaws in the subject strategy. A key weakness of Market Replay is that the simulated market does not substantially adapt to or respond to the presence of the experimental strategy. IABS methods provide an artificial market for the experimental strategy using a population of background trading agents. Because the background agents attend to market conditions and current price as part of their strategy, the overall market is responsive to the presence of the experimental strategy. Even so, IABS methods have their own weaknesses, primarily that it is unclear if the market environment they provide is realistic. We describe our approach in detail, and illustrate its use in an example application: The evaluation of market impact for various size orders.

@misc{https://doi.org/10.48550/arxiv.2004.13152,

doi = {10.48550/ARXIV.2004.13152},

url = {https://arxiv.org/abs/2004.13152},

author = {Lockhart, Joshua and Assefa, Samuel and Balch, Tucker and Veloso, Manuela},

keywords = {Machine Learning (cs.LG), Machine Learning (stat.ML), FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Some people aren't worth listening to: periodically retraining classifiers with feedback from a team of end users},

publisher = {arXiv},

year = {2020},

copyright = {arXiv.org perpetual, non-exclusive license}

}

Document classification is ubiquitous in a business setting, but often the end users of a classifier are engaged in an ongoing feedback-retrain loop with the team that maintain it. We consider this feedback-retrain loop from a multi-agent point of view, considering the end users as autonomous agents that provide feedback on the labelled data provided by the classifier. This allows us to examine the effect on the classifier’s performance of unreliable end users who provide incorrect feedback. We demonstrate a classifier that can learn which users tend to be unreliable, filtering their feedback out of the loop, thus improving performance in subsequent iterations.

@misc{https://doi.org/10.48550/arxiv.1911.05892,

doi = {10.48550/ARXIV.1911.05892},

url = {https://arxiv.org/abs/1911.05892},

author = {Ganesh, Sumitra and Vadori, Nelson and Xu, Mengda and Zheng, Hua and Reddy, Prashant and Veloso, Manuela},

keywords = {Trading and Market Microstructure (q-fin.TR), Machine Learning (cs.LG), Multiagent Systems (cs.MA), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Reinforcement Learning for Market Making in a Multi-agent Dealer Market},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

Financial markets contain a rich set of multiagent learning problems but lack simulators that could be used to research, develop and test reinforcement learning algorithms. In this paper, we demonstrate the use of agent based modeling to simulate a dealer market with two types of agents - market makers and investors. In particular, we focus on the dynamics that arise from price differentiation and risk management. We show through experiments that our simulation model is able to produce known effects in these markets such as varied price sensitivity among investors and the benefits of internalization for market makers.

@article{https://doi.org/10.48550/arxiv.2004.06565,

doi = {10.48550/ARXIV.2004.06565},

url = {https://arxiv.org/abs/2004.06565},

author = {Nagpal, Chirag and Tillman, Robert E. and Reddy, Prashant and Veloso, Manuela},

keywords = {Statistical Finance (q-fin.ST), Machine Learning (cs.LG), Applications (stat.AP), Machine Learning (stat.ML), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Bayesian Consensus: Consensus Estimates from Miscalibrated Instruments under Heteroscedastic Noise},

 publisher = {arXiv}

year = {2020},

copyright = {arXiv.org perpetual, non-exclusive license}

}

We consider the problem of aggregating predictions or measurements from a set of human forecasters, models, sensors or other instruments which may be subject to bias or miscalibration and random heteroscedastic noise. We propose a Bayesian consensus estimator that adjusts for miscalibration and noise and show that this estimator is unbiased and asymptotically more efficient than naive alternatives. We further propose a Hierarchical Bayesian Model that leverages our proposed estimator and apply it to two real world forecasting challenges that require consensus estimates from error prone individual estimates: forecasting influenza like illness (ILI) weekly percentages and forecasting annual earnings of public companies. We demonstrate that our approach is effective at mitigating bias and error and results in more accurate forecasts than existing consensus models.

@misc{https://doi.org/10.48550/arxiv.1911.12816,

doi = {10.48550/ARXIV.1911.12816},

url = {https://arxiv.org/abs/1911.12816},

author = {Mahfouz, Mahmoud and Filos, Angelos and Chtourou, Cyrine and Lockhart, Joshua and Assefa, Samuel and Veloso, Manuela and Mandic, Danilo and Balch, Tucker},

keywords = {Computational Finance (q-fin.CP), Trading and Market Microstructure (q-fin.TR), FOS: Economics and business, FOS: Economics and business},

 title = {On the Importance of Opponent Modeling in Auction Markets}

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

The dynamics of financial markets are driven by the interactions between participants, as well as the trading mechanisms and regulatory frameworks that govern these interactions. Decision makers would rather not ignore the impact of other participants on these dynamics, and should employ tools and models that take this into account. To this end, we demonstrate the efficacy of applying opponent-modeling in a number of simulated market settings. While our simulations are simplified representations of actual market dynamics, they provide an idealized “playground” in which our techniques can be demonstrated and tested. We present this work with the aim that our techniques could be refined and, with some effort, scaled up to the full complexity of real-world market scenarios. We hope that the results presented encourage practitioners to adopt opponent-modeling methods and apply them on live systems, in order to enable not only reactive but also proactive decisions to be made.

@inproceedings{10.1145/3383455.3422554,

author = {Assefa, Samuel A. and Dervovic, Danial and Mahfouz, Mahmoud and Tillman, Robert E. and Reddy, Prashant and Veloso, Manuela},

title = {Generating Synthetic Data in Finance: Opportunities, Challenges and Pitfalls},

year = {2020},

isbn = {9781450375849},

publisher = {Association for Computing Machinery},

address = {New York, NY, USA},

url = {https://doi.org/10.1145/3383455.3422554},

doi = {10.1145/3383455.3422554},

booktitle = {Proceedings of the First ACM International Conference on AI in Finance},

articleno = {44},

numpages = {8},

keywords = {synthetic data, privacy preserving data generation, simulation},

location = {New York, New York},

series = {ICAIF '20}

}

Financial services generate a huge volume of data that is extremely complex and varied. These datasets are often stored in silos within organisations for various reasons, including but not limited to, regulatory requirements and business needs. As a result, data sharing within different lines of business as well as outside of the organisation (e.g. to the research community) is severely limited. It is therefore critical to investigate methods for synthesising financial datasets that follow the same properties of the real data while respecting the need for privacy of the parties involved in a particular dataset. This introductory paper aims to highlight the growing need for effective synthetic data generation in the financial domain. We highlight three main areas of focus for the academic community: 1) Generating realistic synthetic datasets. 2) Measuring the similarities between real and generated datasets 3) Ensuring the generative process satisfies any privacy constraints. Although these challenges are also present in other domains, the extra regulatory and privacy requirements add another dimension of complexity and offer a unique opportunity to study the topic in financial services. Finally, we aim to develop a shared vocabulary and context for generating synthetic financial data using two types of financial datasets as examples.

@misc{https://doi.org/10.48550/arxiv.2010.05867,

doi = {10.48550/ARXIV.2010.05867},

url = {https://arxiv.org/abs/2010.05867},

author = {Byrd, David and Polychroniadou, Antigoni},

keywords = {Cryptography and Security (cs.CR), Artificial Intelligence (cs.AI), Multiagent Systems (cs.MA), General Finance (q-fin.GN), FOS: Computer and information sciences, FOS: Computer and information sciences, FOS: Economics and business, FOS: Economics and business},

title = {Differentially Private Secure Multi-Party Computation for Federated Learning in Financial Applications},

publisher = {arXiv},

year = {2020},

copyright = {arXiv.org perpetual, non-exclusive license}

}

Federated Learning is a technique that enables a large number of users to jointly learn a shared machine learning model, managed by a centralized server, while the training data remains on user devices. That said, federated learning reduces data privacy risks. Privacy concerns still exist since it is possible to leak information about the training data set from the trained model’s weights or parameters. Most federated learning systems use the technique of differential privacy to add noise to the weights so that it is harder to reverse-engineer the individual data sets. Differential privacy reduces the risk but does not eliminate leakage from the data. The combination of differential privacy and cryprogrpahy can eliminate leakage. As opposed to prior works, we propose a new mechanism that protects against a wide range of attacks. Our mechanism is based on advanced cryptographic techniques, in particular, secure multiparty computation and differential privacy. Our model has been developed and tested on the ABIDES environment simulating mobile device networks.

Exchange traded funds (ETF), which are baskets of securities that trade on the stock market, have become a popular investment tool, but potentially increase the price co-movement of their underlying securities. We study index-based ETFs, because underlying securities trade on the same market as their corresponding ETF, which creates ample opportunities for traders to arbitrage between price deviations in the ETF and the market index. Considering the increase in market volatility in 2018 and 2019 in the US stock market, we examine if an ETF spreads mini flash crashes, or small volatility events, from one of its underlying symbols to another. We address this question in a simulated environment with an ETF and two symbols which compose the ETF’s portfolio. We explore two market environments, one with an ETF and arbitrage agents, and one without an ETF and arbitrage agents. We find that the presence of an ETF and arbitrage agents increases underlying symbols’ overall volatility, and when one symbol experiences a mini flash crash, the other symbol does experience a momentary price change in the opposite direction. This supports the idea that ETFs potentially increase market-wide volatility by spreading volatility events through their portfolios. To further this work, we hope to utilize learning to implement smarter arbitrage agents and smarter background agents who might avoid adverse selection by not trading during mini flash crashes.

@unknown{unknown,

author = {Ravi, Vineeth and Amrouni, Selim and Stefanucci, Andrea and Reddy, Prashant and Veloso, Manuela},

year = {2020},

month = {10},

pages = {},

title = {AI pptX: Robust Continuous Learning for Document Generation with AI Insights}

}

Business analysts create billions of slide decks, reports and documents annually. Most of these documents have well-defined structure comprising of similar content generated from data. We present AI pptX, a novel AI framework for creating and modifying documents as well as extract insights in the form of natural language sentences from data. AI pptX has three main components: (i) a component that translates users’ natural language input into “skills” that encapsulate content editing and formatting commands, (ii) a robust continuously learning component that interacts with users, and (iii) a component that automatically generates hierarchical insights in the form of natural language sentences. We illustrate (i) and (ii) with a study of 18 human users tasked to create a presentation deck and observe the learning capability from a decrease in user-input commands by up to 45%. We demonstrate the robust learning capability of AI pptX with experimental simulations of non-collaborative users. We illustrate (i) and (iii) by automatically generating insights in natural language using a data set from the Electricity Transmission Network of France (RTE); we show that a complex statistical analysis of series can automatically be distilled into easily interpretable explanations called AI Insights

@misc{https://doi.org/10.48550/arxiv.1910.13012,

doi = {10.48550/ARXIV.1910.13012},

url = {https://arxiv.org/abs/1910.13012},

author = {Petosa, Nick and Balch, Tucker},

keywords = {Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Multiplayer AlphaZero},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

The AlphaZero algorithm has achieved superhuman performance in two-player, deterministic, zero-sum games where perfect information of the game state is available. This success has been demonstrated in Chess, Shogi, and Go where learning occurs solely through self-play. Many real-world applications (e.g., equity trading) require the consideration of a multiplayer environment. In this work, we suggest novel modifications of the AlphaZero algorithm to support multiplayer environments, and evaluate the approach in two simple 3-player games. Our experiments show that multiplayer AlphaZero learns successfully and consistently outperforms a competing approach: Monte Carlo tree search. These results suggest that our modified AlphaZero can learn effective strategies in multiplayer game scenarios. Our work supports the use of AlphaZero in multiplayer games and suggests future research for more complex environments.

@misc{https://doi.org/10.48550/arxiv.1911.05892,

doi = {10.48550/ARXIV.1911.05892},

url = {https://arxiv.org/abs/1911.05892},

author = {Ganesh, Sumitra and Vadori, Nelson and Xu, Mengda and Zheng, Hua and Reddy, Prashant and Veloso, Manuela},

keywords = {Trading and Market Microstructure (q-fin.TR), Machine Learning (cs.LG), Multiagent Systems (cs.MA), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Reinforcement Learning for Market Making in a Multi-agent Dealer Market},

publisher = {arXiv},

 year = {2019}

copyright = {arXiv.org perpetual, non-exclusive license}

}

Market makers play an important role in providing liquidity to markets by continuously quoting prices at which they are willing to buy and sell, and managing inventory risk. In this paper, we build a multi-agent simulation of a dealer market and demonstrate that it can be used to understand the behavior of a reinforcement learning (RL) based market maker agent. We use the simulator to train an RL-based market maker agent with different competitive scenarios, reward formulations and market price trends (drifts). We show that the reinforcement learning agent is able to learn about its competitor’s pricing policy; it also learns to manage inventory by smartly selecting asymmetric prices on the buy and sell sides (skewing), and maintaining a positive (or negative) inventory depending on whether the market price drift is positive (or negative). Finally, we propose and test reward formulations for creating risk averse RL-based market maker agents.

@article{https://doi.org/10.48550/arxiv.1912.04941,

doi = {10.48550/ARXIV.1912.04941},

url = {https://arxiv.org/abs/1912.04941},

author = {Vyetrenko, Svitlana and Byrd, David and Petosa, Nick and Mahfouz, Mahmoud and Dervovic, Danial and Veloso, Manuela and Balch, Tucker Hybinette},

keywords = {Trading and Market Microstructure (q-fin.TR), Multiagent Systems (cs.MA), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},

title = {Get Real: Realism Metrics for Robust Limit Order Book Market Simulations},

publisher = {arXiv},

year = {2019},

copyright = {arXiv.org perpetual, non-exclusive license}

}

Machine learning (especially reinforcement learning) methods for trading are increasingly reliant on simulation for agent training and testing. Furthermore, simulation is important for validation of hand-coded trading strategies and for testing hypotheses about market structure. A challenge, however, concerns the robustness of policies validated in simulation because the simulations lack fidelity. In fact, researchers have shown that many market simulation approaches fail to reproduce statistics and stylized facts seen in real markets. As a step towards addressing this we surveyed the literature to collect a set of reference metrics and applied them to real market data and simulation output. Our paper provides a comprehensive catalog of these metrics including mathematical formulations where appropriate. Our results show that there are still significant discrepancies between simulated markets and real ones. However, this work serves as a benchmark against which we can measure future improvement.