Research Center for AI Finance and Deep Learning

The Research Center for AI Finance and Deep Learning (abbreviated as “IDEA-FinAI”) is the first core research center in IDEA Research founded by Dr. Harry Shum. Our vision is to build an AI super-brain more intelligent than human being in making financial decisions (abbreviated as “FinAI-X” project). With the rapid growth of AI technology, big data technology and super-computing technology, an unprecedented revolution is sweeping the finance industry. The mission of FinAI is to provide a world-leading platform for all scientists and engineers devoting to the interdisciplinary field of AI finance to encourage and support the state-of-the-art research and innovation, and make contribution to all finance industry. 

The hierarchy of our FinAI-X project consists of five layers: super-computers, data systems, knowledge graphs, decision engines and industrial scenarios/applications (as shown in the figure below). 

Super Computing System: super-computing clusters are hardware infrastructure and play a fundamental role in FinAI-X.  Based on the current CPU-GPU heterogenous computing structure, we are developing new hardware/software techniques to parallelize and accelerate the finance computing, including various finance algorithms, big data algorithms, statistical machine learning algorithms, deep learning algorithms and reinforcement learning algorithms. System accelerating techniques such as distributed computing optimization, system communication optimization, memory computing optimization, instruction set optimization and FPGA have boosted many algorithms up to 1000x in computing performance. 

Data System: through cooperation with third parties, we are building a large-scale data mining and management system for various finance data across all mainstream trading markets, including but not limited to finance quotation data with various frequencies and various resolution, order/limit book data, event behavioral data, financial analyst data, supply chain data, traditional fundamental data, and alternative fundamental data. Our data system comes up with an automatic data processing pipeline and an automatic feature engineering pipeline to collect and analyze full profiles of financial entities (e.g., stocks, futures, options, bonds, enterprises, sectors, key figures and key events) and to analyze their dynamics. Techniques such as distributed storage, low latency I/O and distributed graph databases are developed to improve the performance of data processing. 

Knowledge Graph System: based on financial databases and financial entity profiling, we applied AI techniques such as NLP, deep learning and small sample learning to construct various and diverse relationships among financial entities (industry chain, capital chain, supply chain, litigation chains, etc.) as well as complicated causal relationships between events. In addition, we develop new temporal reasoning algorithms and new techniques that are applicable to large-scale financial behavioral knowledge graph analysis to find the optimal trade-off among prediction accuracy, interpretability, computational complexity, flexibility for distributed computing, confidence level and robustness. 

Decision Engine: based on financial entity profiling and financial behavioral knowledge graphs, we research and develop new techniques to improve accuracies and robustness of prediction tasks in various financial service scenarios, including large-scale spatial-temporal deep learning models with hundreds of billions of parameters, large-scale deep reinforcement learning model, new autoML techniques based on NAS/MoE, large-scale automatic finance signal mining system and more efficient search and optimization algorithms. We are exploring and developing both software and hardware acceleration techniques in order to support the distributed training and low-latency inference for super-large-scale financial deep learning model. 

Industrial Applications: the application layer of FinAI-X focuses on demands and pain points of financial sectors such as security services, mutual/hedge funds, investments banks, commercial banks and insurance services, as well as the strategic demand of financial risk management and security. In the next five years, our research covers AI asset/wealth management, AI investment and risk control, AI cloud pricing system for derivatives in global markets, simulator of securities trading market and stress test techniques, investment research robots, etc. 

Key Research Projects

1.Deep/reinforcement learning problems in financial investment and risk management: 

Analysis of financial time series suffers a number of difficulties, including extremely noisy data (usually signal-to-noise ratios ranging from 1:100 to 1:1000), curse of non-stationarity (especially when the market efficiency is getting higher and higher and the market style is switching faster and faster, resulting in the accelerative decay of finance signals and machine learning models), lack of consistency (the steps of pricing/prediction, positioning/portfolio optimization and execution optimization are computed independently, rather than a global optimization process), and poor interpretability (deep learning models are black boxes and they present obstacles in human understanding, empirical analysis and cognitive reasoning). To address these problems, we are developing new machine/deep/reinforcement learning algorithms as well as new data modeling frameworks to significantly improve the performance of prediction, interpretability and cognitive reasoning. 

2.Large-scale automatic financial signal mining and autoML modeling:

Financial signals (or factors) are a series of mathematical rules, statistical models or machine learning models that characterize certain laws in financial markets and have certain capabilities either in predicting market trends or in monitoring/forecasting market risks. Traditionally financial signals are designed and modeled by investment researchers or quant researchers manually, resulting in low efficiency and high cost in research processes. To address these limitations, we rebuild the investment research procedure using autoML framework. Specifically, we automate the process of data preprocessing, feature engineering, deep learning modeling, execution optimization and risk monitoring/control in a unified end2end pipeline. In the feature engineering step, in particular, we model the financial signal mining process as a search and optimization problem in complicated mathematical expression spaces, and develop new AI algorithms and distributed computing techniques to discover effective financial signals in an efficient and scalable way. 

3.Finance behavioral knowledge graph and cognitive graph reasoning: 

In the next five years, we aim to build a large dynamic knowledge graph collecting various financial and economic behaviors, covering different industrial chains, sectors, markets and geolocations, to mine complex relationships among various finance entities and events, as well as their relationships, causalities and dynamics. Based on the knowledge graph data we are developing, our methodological research covers new symbolic reasoning algorithms, probabilistic graphical model, and distributed graph neural network algorithms to better analyze the financial behaviors collected from data.  

4.Large-scale finance deep learning model for multimodal and multitask learning:

Self-supervised deep learning pre-training models such as BERT, GPT, and DALLE/CLIP have achieved significant success in diverse NLP tasks, and we believe that this normal form has great potential in finance modeling and analysis. We firmly believe that such paradigms have great research and application potentials in the financial field. We train ultra-large-scale multimodal temporal deep learning models with trillions of parameters to learn high-order patterns from global financial markets. We research general AI models for finance scenarios across various markets and resolutions. We study new model frameworks beyond well-known models such as LSTM, Transformer and MLP-mixer etc., and we develop new autoML techniques/pipelines to improve the efficiency of ultra-large-scale model training and inference. 

5.Software and hardware acceleration for financial algorithms: 

With the rapid development of algorithmic technology, super-computing technology and data technology, finance data analysis is evolving towards ultrafine data granularity, multiple data resolution and diversified data sources/dimensions, and therefore it catalyzes rigid demand for large model computing acceleration at algorithm level, software level and hardware level, simultaneously. We study CPU-GPU-FPGA heterogenous computing acceleration, FPGA extremely low latency computing, as well as distributed computing acceleration for classic deep learning models such as Transformer, GCN/GAT and classic model optimization/ensemble techniques such as NAS/MoE.

6.AI model and cloud computing system for derivatives pricing:  

Financial derivatives such as stock options, index options and interest rate swaps have huge volumes in global financial markets. Derivatives investment and risk management rely mainly on the accurate computation of implied volatility, which plays a key role in pricing. Traditionally prices and implied volatilities are computed by solving complex mathematical models such as stochastic differential equations, and therefore the pricing accuracy depends heavily on the correctness of model assumptions as well as the effectiveness of ODE/PDE solvers. Our research aims to utilize advantages of deep learning in learning and prediction as well as advantages of ODE/PDE in model interpretability, and to provide more accurate pricing computation and related cloud computing services. 

The Team

IDEA-FinAI is led by senior experts in the interdisciplinary field of AI finance, including an ex-Harvard data science professor. The team of FinAI has many brilliant young scientists/engineers sharing the common mission to revolutionize financial industry with cutting-edge AI technology. Our colleagues graduated from many globally renowned universities, such as Harvard, Columbia, UC-Berkeley, Cornell, Michigan, Northwestern, UIUC, and Cambridge, and they had work experiences in famous IT or finance companies. We seek talents with “scientific mind, engineering persistence and pioneering spirit”, and we create and maintain an exciting engineering culture and work atmosphere with innovation, sharing, fairness and achievement. We believe every genius young scientist/engineer has unlimited potential and creativity. Let’s get together to challenge the cutting-edge technical problems and to change the finance industry, improving the fairness, efficiency, effectiveness and security.  

FinAI contact information: