Hung-Ching Chen

Inferring agent dynamics from social communication network

We present a machine learning approach to discovering the agent dynamics or micro-laws that drives the evolution of the social groups in a community. We set up the problem by introducing a parameterized probabilistic model for the agent dynamics: the acts of an agent are determined by micro-laws with unknown parameters. Our approach is to identify the appropriate micro-laws which corresponds to identifying the appropriate parameters in the model. To solve the problem we develop heuristic expectation-maximization style algorithms for determining the micro-laws of a community based on either the observed social group evolution, or observed set of communications between actors. We present the results of extensive experiments on simulated data as well as some results on real communities, e.g., newsgroups.

Reverse engineering a social agent-based hidden markov model--visage.

We present a machine learning approach to discover the agent dynamics that drives the evolution of the social groups in a community. We set up the problem by introducing an agent-based hidden Markov model for the agent dynamics: an agents actions are determined by micro-laws. Nonetheless, We learn the agent dynamics from the observed communications without knowing state transitions. Our approach is to identify the appropriate micro-laws corresponding to an identification of the appropriate parameters in the model. The model identification problem is then formulated as a mixed optimization problem. To solve the problem, we develop a multistage learning process for determining the group structure, the group evolution, and the micro-laws of a community based on the observed set of communications among actors, without knowing the semantic contents. Finally, to test the quality of our approximations and the feasibility of the approach, we present the results of extensive experiments on synthetic data as well as the results on real communities, such as Enron email and Movie newsgroups. Insight into agent dynamics helps us understand the driving forces behind social evolution.

Identifying Multi-ID Users in Open Forums

We describe a model for real-time communication exchange in public forums, such as newsgroups and chatrooms, and use this model to develop an efficient algorithm which identifies the users that post their messages under different IDs, multi-ID users. Our simulations show that, under the model’s assumptions, the identification of multi-ID users is highly effective, with false positive and false negative rates of about 0.1% in the worst case.

The Multilevel Classification Problem and a Monotonicity Hint

We introduce and formalize the multilevel classification problem, in which each category can be subdivided into different levels. We analyze the framework in a Bayesian setting using Normal class conditional densities. Within this framework, a natural monotonicity hint converts the problem into a nonlinear programming task, with non-linear constraints. We present Monte Carlo and gradient based techniques for addressing this task, and show the results of simulations. Incorporation of monotonicity yields a systematic improvement in performance.

Reverse engineering an agent-based hidden Markov model for complex social systems

The power of social values that helps to shape or formulate our behavior patterns is not only inevitable, but also how we have surreptitiously responded to the hidden curriculum that derives from such social values in our decision making can be just as significant. Through a machine learning approach, we are able to discover the agent dynamics that drives the evolution of the social groups in a community. By doing so, we set up the problem by introducing an agent-based hidden Markov model, in which the acts of an agent are determined by microlaws with unknown parameters. To solve the problem, we develop a multistage learning process for determining the micro-laws of a community based on observed set of communications between actors without the semantic contents. We present the results of extensive experiments on synthetic data as well as some results on real communities, e.g., Enron email and movie newsgroups.

Discover the power of social and hidden curriculum to decision making: experiments with enron email and movie newsgroups

The power of social values that helps to surreptitiously shape or formulate our behavior patterns is not only inevitable, but also influential as the directions of our decision making can never seem to escape the impact of this hidden agent. Therefore, the search of such power agent can be validated through a machine learning approach that enables us to discover the agent dynamics in which drives the evolution of the social groups in a community. By doing so, we set up the problem by introducing a parameterized probabilistic model for the agent dynamics: the acts of an agent are determined by micro-laws with unknown parameters. Our approach is to identify the appropriate parameters in the model. To solve the problem, we develop heuristic expectation-maximization style algorithms for determining the micro-laws of a community based on either observed communication links between actors, or the observed evolution of social groups. We present the learning results from the synthetic data as well as the findings on real communities, e.g., Enron email and movie newsgroups.

NN-OPT: neural network for option pricing using multinomial tree

We provide a framework for learning to price complex options by learning risk-neutral measures (Martingale measures). In a simple geometric Brownian motion model, the price volatility, fixed interest rate and a no-arbitrage condition suffice to determine a unique riskneutral measure. On the other hand, in our framework, we relax some of these assumptions to obtain a class of allowable risk-neutral measures. We then propose a framework for learning the appropriate risk-neural measure. In particular, we provide an efficient algorithm for backpropagating gradients through multinomial pricing trees. Since the risk-neutral measure prices all options simultaneously, we can use all the option contracts on a particular stock for learning.We demonstrate the performance of these models on historical data. Finally, we illustrate the power of such a framework by developing a real time trading system based upon these pricing methods.

ViSAGE: A Vi rtual Laboratory for S imulation and A nalysis of Social G roup E volution

We present a modeling laboratory, Virtual Laboratory for the Simulation and Analysis of Social Group Evolution (ViSAGE), that views the organization of human communities and the experience of individuals in a community as contingent upon on the dynamic properties, or micro-laws, of social groups. The laboratory facilitates the theorization and validation of these properties through an iterative research processes that involves (1) forward simulation experiments, which are used to formalize dynamic group properties, (2) reverse engineering from real data on how the parameters are distributed among individual actors in the community, and (3) grounded research, such as participant observation, that follows specific activities of real actors in a community and determines if, or how well, the micro-laws describe the way choices are made in real world, local settings. In this article we report on the design of ViSAGE. We first give some background to the model. Next we detail each component. We then describe a set of simulation experiments that we used to further design and clarify ViSAGE as a tool for studying emergent properties/phenomena in social networks.

Learning Martingale Measures From High Frequency Financial Data to Help Option Pricing

We provide a framework for learning risk-neutral measures (Martingale measures) for pricing options from high frequency financial data. In a simple geometric Brownian motion model, a price volatility, a fixed interest rate and a no-arbitrage condition suffice to determine a unique risk-neutral measure. On the other hand, in our framework, we relax some of these assumptions to obtain a class of allowable risk-neutral measures. We then propose a framework for learning the appropriate risk-neural measure. Since the riskneutral measure prices all options simultaneously, we can use all the option contracts on a particular underlying stock for learning. We demonstrate the performance of these models on historical data. In particular, we show that both learning without a no-arbitrage condition and a no-arbitrage condition without learning are worse than our framework; however the combination of learning with a no-arbitrage condition has the best result. These results indicate the potential to learn Martingale measures with a no-arbitrage condition providing just the right constraint. We also compare our approach to standard Binomial models with volatility estimates (historical volatility and GARCH volatility predictors). Finally, we illustrate the power of such a framework by developing a real time trading system based upon these pricing methods.