Inference time latency has remained an open challenge for real world applications of large language models (LLMs). State-of-the-art (SOTA) speculative sampling (SpS) methods for LLMs, like EAGLE-3, use tree-based drafting to explore multiple candidate continuations in parallel. However, the hyperparameters controlling the tree structure are static, which limits flexibility and efficiency across diverse contexts and domains. We introduce Reinforcement learning for Speculative Sampling (Re-SpS), the first reinforcement learning (RL)-based framework for draft tree hyperparameter optimization. Re-SpS dynamically adjusts draft tree hyperparameters in real-time, learning context-aware policies that maximize generation speed by balancing speculative aggression with computational overhead. It leverages efficient state representations from target model hidden states and introduces multi-step action persistence for better context modeling. Evaluation results across five diverse benchmarks demonstrate consistent improvements over the SOTA method EAGLE-3, achieving up to 5.45x speedup over the backbone LLM and up to 1.12x speedup compared to EAGLE-3 across five diverse benchmarks, with no loss in output fidelity.

Diffusion models have shown promising ability in generating high-quality time series (TS) data. Despite the initial success, existing works mostly focus on the authenticity of data at the individual level, but pay less attention to preserving the population-level properties on the entire dataset. Such population-level properties include value distributions for each dimension and distributions of certain functional dependencies (e.g., cross-correlation, CC) between different dimensions. For instance, when generating house energy consumption TS data, the value distributions of the outside temperature and the kitchen temperature should be preserved, as well as the distribution of CC between them. Preserving such TS population-level properties is critical in maintaining the statistical insights of the datasets, mitigating model bias, and augmenting downstream tasks like TS prediction. Yet, it is often overlooked by existing models. Hence, data generated by existing models often bear distribution shifts from the original data. We propose Population-aware Diffusion for Time Series (PaD-TS), a new TS generation model that better preserves the population-level properties. The key novelties of PaD-TS include 1) a new training method explicitly incorporating TS population-level property preservation, and 2) a new dual-channel encoder model architecture that better captures the TS data structure. Empirical results in major benchmark datasets show that PaD-TS can improve the average CC distribution shift score between real and synthetic data by 5.9x while maintaining a performance comparable to state-of-the-art models on individual-level authenticity.

Natural Language to SQL (NL2SQL) enables natural language access to structured data, but LLM-based methods can be inefficient for real-time use and repetitive query patterns. We present AutoRuleSQL, a hybrid system that combines template-based fast paths with LLM fallback and offline bootstrapping. Empirical results show that it reduces latency by over 12.6% and improves execution accuracy by up to 4.0%, when combined with existing NL2SQL methods.

Large language models (LLMs) excel at factual recall yet still propagate stale or incorrect knowledge.In-context knowledge editing offers a gradient-free remedy suitable for black-box APIs, but current editors rely on static demonstration sets chosen by surface-level similarity, leading to two persistent obstacles: (i) a quantity-quality trade-off, and (ii) lack of adaptivity to task difficulty.We address these issues by dynamically selecting supporting demonstrations according to their utility for the edit.We propose Dynamic Retriever for In-Context Knowledge Editing (DR-IKE), a lightweight framework that (1) trains a BERT retriever with REINFORCE to rank demonstrations by editing the reward, and (2) employs a learnable threshold to prune low-value examples, shortening the prompt when the edit is easy and expanding it when the task is hard.DR-IKE performs editing without modifying model weights, relying solely on forward passes for compatibility with black-box LLMs.On the COUNTERFACT benchmark, it improves edit success by up to 17.1%, reduces latency by 41.6%, and preserves accuracy on unrelated queries, demonstrating scalable and adaptive knowledge editing.

Reinforcement learning (RL) has emerged as a promising tool for combinatorial optimization (CO) problems due to its ability to learn fast, effective, and generalizable solutions. Nonetheless, existing works mostly focus on one-shot deterministic CO, while sequential stochastic CO (SSCO) has rarely been studied despite its broad applications such as adaptive influence maximization (IM) and infectious disease intervention. In this paper, we study the SSCO problem where we first decide the budget (e.g., number of seed nodes in adaptive IM) allocation for all time steps, and then select a set of nodes for each time step. The few existing studies on SSCO simplify the problems by assuming a uniformly distributed budget allocation over the time horizon, yielding suboptimal solutions. We propose a generic hierarchical RL (HRL) framework called wake-sleep option (WS-option), a two-layer option-based framework that simultaneously decides adaptive budget allocation on the higher layer and node selection on the lower layer. WS-option starts with a coherent formulation of the two-layer Markov decision processes (MDPs), capturing the interdependencies between the two layers of decisions. Building on this, WS-option employs several innovative designs to balance the model's training stability and computational efficiency, preventing the vicious cyclic interference issue between the two layers. Empirical results show that WS-option exhibits significantly improved effectiveness and generalizability compared to traditional methods. Moreover, the learned model can be generalized to larger graphs, which significantly reduces the overhead of computational resources.

There have been extensive studies on learning in zero-sum games, focusing on the analysis of the existence and algorithmic convergence of Nash equilibrium (NE). Existing studies mainly focus on symmetric games where the strategy spaces of the players are of the same type and size. For the few studies that do consider asymmetric games, they are mostly restricted to matrix games. In this paper, we define and study a new practical class of asymmetric games called two-player Asymmetric Combinatorial-Continuous zEro-Sum (ACCES) games, featuring a combinatorial action space for one player and an infinite compact space for the other. Such ACCES games have broad implications in the real world, particularly in combinatorial optimization problems (COPs) where one player optimizes a solution in a combinatorial space, and the opponent plays against it in an infinite (continuous) compact space (e.g., a nature player deciding epistemic parameters of the environmental model). Our first key contribution is to prove the existence of NE for two-player ACCES games, using the idea of essentially finite game approximation. Building on the theoretical insights and double oracle (DO)-based solutions to complex zero-sum games, our second contribution is to design the novel algorithm, Combinatorial Continuous DO (CCDO), to solve ACCES games, and prove the convergence of the proposed algorithm. Considering the NP-hardness of most COPs and recent advancements in reinforcement learning (RL)-based solutions to COPs, our third contribution is to propose a practical algorithm to solve NE in the real world, CCDORL (based on CCDO) and provide the novel convergence analysis in the ACCES game. Experimental results across diverse instances of COPs demonstrate the empirical effectiveness of our algorithms.

Large language models (LLMs) have transformed automated code generation. However, their high computational demands often lead to server overload and increased latency in SaaS deployments. To address this, we present SpeCoder, a framework that accelerates server-side code generation using speculative sampling (SpS) and supervised fine-tuning (SFT). SpS allows lower latency in the code generation, whereas SFT enables more personalized code generation tailored to the user's needs.

Learning expressive stochastic policies instead of deterministic ones has been proposed to achieve better stability, sample complexity and robustness. Notably, in Maximum Entropy reinforcement learning (MaxEnt RL), the policy is modeled as an expressive energy-based model (EBM) over the Q-values. However, this formulation requires the estimation of the entropy of such EBM distributions which is an open problem. To address this, previous MaxEnt RL methods either implicitly estimate the entropy, yielding high computational complexity and variance (SQL), or follow a variational inference approach that fits simplified distributions (e.g., Gaussian) for tractability (SAC). We propose Sein Soft Actor-Critic (S$^2$AC), a MaxEnt RL algorithm that learns expressive policies without compromising efficiency. S$^2$AC uses parameterized Stein Variational Gradient Descent (SVGD) as the underlying policy. At the core of S$^2$AC is a new solution to the above open challenge of entropy computation for EBMs. Our entropy formula is computationally efficient and only depends on first-order derivatives and vector products. Empirical results show that S$^2$AC yields more optimal solutions to the MaxEnt objective than SQL and SAC in the multi-goal environment, and outperforms SAC and SQL on the MuJoCo benchmark. Our code is available at: https://anonymous.4open.science/r/Stein-Soft-Actor-Critic/

We study an online learning problem in general-sum Stackelberg games, where players act in a decentralized and strategic manner. We study two settings depending on the type of information for the follower: (1) the limited information setting where the follower only observes its own reward, and (2) the side information setting where the follower has extra side information about the leader’s reward. We show that for the follower, myopically best responding to the leader’s action is the best strategy for the limited information setting, but not necessarily so for the side information setting – the follower can manipulate the leader’s reward signals with strategic actions, and hence induce the leader’s strategy to converge to an equilibrium that is better off for itself. Based on these insights, we study decentralized online learning for both players in the two settings. Our main contribution is to derive last iterate convergence and sample complexity results in both settings. Notably, we design a new manipulation strategy for the follower in the latter setting, and show that it has an intrinsic advantage against the best response strategy. Our theories are also supported by empirical results.

In influence maximization (IM), the goal is to find a set of seed nodes in a social network that maximizes the influence spread. While most IM problems focus on classical influence cascades (e.g., Independent Cascade and Linear Threshold) which assume individual influence cascade probability is independent of the number of neighbors, recent studies by sociologists show that many influence cascades follow a pattern called complex contagion (CC), where influence cascade probability is much higher when more neighbors are influenced. Nonetheless, there are very limited studies for complex contagion influence maximization (CCIM) problems. This is partly because CC is non-submodular, the solution of which has been an open challenge. In this study, we propose the first reinforcement learning (RL) approach to CCIM. We find that a key obstacle in applying existing RL approaches to CCIM is the reward sparseness issue, which comes from two distinct sources. We then design a new RL algorithm that uses the CCIM problem structure to address the issue. Empirical results show that our approach achieves the state-of-the-art performance on 9 real-world networks.

Identifying persuasive speakers in an adversarial environment is a critical task. In a national election, politicians would like to have persuasive speakers campaign on their behalf. When a company faces adverse publicity, they would like to engage persuasive advocates for their position in the presence of adversaries who are critical of them. Debates represent a common platform for these forms of adversarial persuasion. This paper solves two problems: the Debate Outcome Prediction (DOP) problem predicts who wins a debate while the Intensity of Persuasion Prediction (IPP) problem predicts the change in the number of votes before and after a speaker speaks. Though DOP has been previously studied, we are the first to study IPP. Past studies on DOP fail to leverage two important aspects of multimodal data: 1) multiple modalities are often semantically aligned, and 2) different modalities may provide diverse information for prediction. Our$\mathsf{M2P2}$(Multimodal Persuasion Prediction) framework is the first to use multimodal (acoustic, visual, language) data to solve the IPP problem. To leverage the alignment of different modalities while maintaining the diversity of the cues they provide,$\mathsf{M2P2}$devises a novel adaptive fusion learning framework which fuses embeddings obtained from two modules – analignmentmodule that extracts shared information between modalities and aheterogeneitymodule that learns the weights of different modalities with guidance from three separately trained unimodal reference models. We test$\mathsf{M2P2}$on the popular IQ2US dataset designed for DOP. We also introduce a new dataset called QPS (from Qipashuo, a popular Chinese debate TV show) for IPP.$\mathsf{M2P2}$significantly outperforms 4 recent baselines on both datasets.

Improving health equity is an urgent task for our society. The advent of mobile clinics plays an important role in enhancing health equity, as they can provide easier access to preventive healthcare for patients from disadvantaged populations. For effective functioning of mobile clinics, accurate prediction of demand (expected number of individuals visiting mobile clinic) is the key to their daily operations and staff/resource allocation. Despite its importance, there have been very limited studies on predicting demand of mobile clinics. To the best of our knowledge, we are among the first to explore this area, using AI-based techniques. A crucial challenge in this task is that there are no known existing data sources from which we can extract useful information to account for the exogenous factors that may affect the demand, while considering protection of client privacy. We propose a novel methodology that completely uses public data sources to extract the features, with several new components that are designed to improve the prediction. Empirical evaluation on a real-world dataset from the mobile clinic The Family Van shows that, by leveraging publicly available data (which introduces no extra monetary cost to the mobile clinics), our AI-based method achieves 26.4% - 51.8% lower Root Mean Squared Error (RMSE) than the historical average-based estimation (which is presently employed by mobile clinics like The Family Van). Our algorithm makes it possible for mobile clinics to plan proactively, rather than reactively, as what has been doing.

Micronutrient deficiency (MND), which is a form of malnutrition that can have serious health consequences, is difficult to diagnose in early stages without blood draws, which are expensive and time-consuming to collect and process. It is even more difficult at a public health scale seeking to identify regions at higher risk of MND. To provide data more widely and frequently, we propose an accurate, scalable, low-cost, and interpretable regional-level MND prediction system. Specifically, our work is the first to use satellite data, such as forest cover, weather, and presence of water, to predict deficiency of micronutrients such as iron, Vitamin B12, and Vitamin A, directly from their biomarkers. We use real-world, ground truth biomarker data collected from four different regions across Madagascar for training, and demonstrate that satellite data are viable for predicting regional-level MND, surprisingly exceeding the performance of baseline predictions based only on survey responses. Our method could be broadly applied to other countries where satellite data are available, and potentially create high societal impact if these predictions are used by policy makers, public health officials, or healthcare providers.

In this work we consider the problem of how to best allocate a limited supply of vaccines in the aftermath of an infectious disease outbreak by viewing the problem as a sequential game between a learner and an environment (specifically, a bandit problem). The difficulty of this problem lies in the fact that the payoff of vaccination cannot be directly observed, making it difficult to compare the relative effectiveness of vaccination on different population groups. Currently used vaccination policies make recommendations based on mathematical modelling and ethical considerations. These policies are static, and do not adapt as conditions change. Our aim is to design and evaluate an algorithm which can make use of routine surveillance data to dynamically adjust its recommendation. We evaluate the performance of our approach by applying it to a simulated epidemic of a disease based on real-world COVID-19 data, and show that our vaccination policy was able to perform better than existing vaccine allocation policies. In particular, we show that with our allocation method, we can reduce the number of required vaccination by at least 50% in order to keep the peak number of hospitalised patients below a certain threshold. Also, when the same batch sizes are used, our method can reduce the peak number of hospitalisation by up to 20%. We also demonstrate that our vaccine allocation does not vary the number of batches per group much, making it socially more acceptable (as it reduces uncertainty, hence results in better and more interpretable communication).

We propose PCAM , a Probabilistic Cyber-Alert Management framework, that enables chief information security officers to better manage cyber-alerts. Workers in Cyber Security Operation Centers usually work in 8- or 12-hour shifts. Before a shift, PCAM analyzes data about all past alerts and true alerts during the shift time-frame to schedule a given set of analysts in accordance with workplace constraints so that the expected number of “uncovered” true alerts (i.e., true alerts not shown to an analyst) is minimized. PCAM achieves this by formulating the problem as a bi-level non-linear optimization problem and then shows how to linearize and solve this complex problem. We have tested PCAM extensively. Using statistics derived from 44 days of real-world alert data, we are able to minimize the expected number of true alerts that are not manually examined by a team consisting of junior, senior, and principal analysts. We are also able to identify the optimal mix of junior, senior, and principal analysts needed during both day and night shifts given a budget, outperforming some reasonable baselines. We tested PCAM ’s proposed schedule (from statistics on 44 days) on a further 6 days of data, using an off-the-shelf false alarm classifier to predict which alerts are real and which ones are false. Moreover, we show experimentally that PCAM is robust to various kinds of errors in the statistics used.

As one of the most powerful topic models, Latent Dirichlet Allocation (LDA) has been used in a vast range of tasks, including document understanding, information retrieval and peer-reviewer assignment. Despite its tremendous popularity, the security of LDA has rarely been studied. This poses severe risks to security-critical tasks such as sentiment analysis and peer-reviewer assignment that are based on LDA. In this paper, we are interested in knowing whether LDA models are vulnerable to adversarial perturbations of benign document examples during inference time. We formalize the evasion attack to LDA models as an optimization problem and prove it to be NP-hard. We then propose a novel and efficient algorithm, EvaLDA to solve it. We show the effectiveness of EvaLDA via extensive empirical evaluations. For instance, in the NIPS dataset, EvaLDA can averagely promote the rank of a target topic from 10 to around 7 by only replacing 1% of the words with similar words in a victim document. Our work provides significant insights into the power and limitations of evasion attacks to LDA models.

In the predict-then-optimize framework, the objective is to train a predictive model, mapping from environment features to parameters of an optimization problem, which maximizes decision quality when the optimization is subsequently solved. Recent work on decision-focused learning shows that embedding the optimization problem in the training pipeline can improve decision quality and help generalize better to unseen tasks compared to relying on an intermediate loss function for evaluating prediction quality. We study the predict-then-optimize framework in the context of sequential decision problems (formulated as MDPs) that are solved via reinforcement learning. In particular, we are given environment features and a set of trajectories from training MDPs, which we use to train a predictive model that generalizes to unseen test MDPs without trajectories. Two significant computational challenges arise in applying decision-focused learning to MDPs: (i) large state and action spaces make it infeasible for existing techniques to differentiate through MDP problems, and (ii) the high-dimensional policy space, as parameterized by a neural network, makes differentiating through a policy expensive. We resolve the first challenge by sampling provably unbiased derivatives to approximate and differentiate through optimality conditions, and the second challenge by using a low-rank approximation to the high-dimensional sample-based derivatives. We implement both Bellman-based and policy gradient-based decision-focused learning on three different MDP problems with missing parameters, and show that decision-focused learning performs better in generalization to unseen tasks.

The influence maximization (IM) problem aims at finding a subset of seed nodes in a social network that maximize the spread of influence. In this study, we focus on a sub-class of IM problems, where whether the nodes are willing to be the seeds when being invited is uncertain, called contingency-aware IM. Such contingency aware IM is critical for applications for non-profit organizations in low resource communities (e.g., spreading awareness of disease prevention). Despite the initial success, a major practical obstacle in promoting the solutions to more communities is the tremendous runtime of the greedy algorithms and the lack of high performance computing (HPC) for the non-profits in the field – whenever there is a new social network, the non-profits usually do not have the HPCs to recalculate the solutions. Motivated by this and inspired by the line of works that use reinforcement learning (RL) to address combinatorial optimization on graphs, we formalize the problem as a Markov Decision Process (MDP), and use RL to learn an IM policy over historically seen networks, and generalize to unseen networks with negligible runtime at test phase. To fully exploit the properties of our targeted problem, we propose two technical innovations that improve the existing methods, including state-abstraction and theoretically grounded reward shaping. Empirical results show that our method achieves influence as high as the state-of-the-art methods for contingency-aware IM, while having negligible runtime at test phase.

Green security domains feature defenders who plan patrols in the face of uncertainty about the adversarial behavior of poachers, illegal loggers, and illegal fishers. Importantly, the deterrence effect of patrols on adversaries’ future behavior makes patrol planning a sequential decision-making problem. Therefore, we focus on robust sequential patrol planning for green security following the minimax regret criterion, which has not been considered in the literature. We formulate the problem as a game between the defender and nature who controls the parameter values of the adversarial behavior and design an algorithm MIRROR to find a robust policy. MIRROR uses two reinforcement learning–based oracles and solves a restricted game considering limited defender strategies and parameter values. We evaluate MIRROR on real-world poaching data.

Human feedback is widely used to train agents in many domains. However, previous works rarely consider the uncertainty when humans provide feedback, especially in cases that the optimal actions are not obvious to the trainers. For example, the reward of a sub-optimal action can be stochastic and sometimes exceeds that of the optimal action, which is common in games or real-world. Trainers are likely to provide positive feedback to sub-optimal actions, negative feedback to the optimal actions and even do not provide feedback in some confusing situations. Existing works, which utilize the Expectation Maximization (EM) algorithm and treat the feedback model as hidden parameters, do not consider uncertainties in the learning environment and human feedback. To address this challenge, we introduce a novel feedback model that considers the uncertainty of human feedback. However, this incurs intractable calculus in the EM algorithm. To this end, we propose a novel approximate EM algorithm, in which we approximate the expectation step with the Gradient Descent method. Experimental results in both synthetic scenarios and two real-world scenarios with human participants demonstrate the superior performance of our proposed approach.

Although most game theory models assume that payoff matrices are provided as input, getting payoff matrices in strategic games (e.g., corporate negotiations and counter-terrorism operations) has proven difficult. To tackle this challenge, we propose a payoff inference engine (PIE) that finds payoffs assuming that players in a game follow a myopic best response or a regret minimization heuristic. This assumption yields a set of constraints (possibly nonlinear) on the payoffs with a multiplicity of solutions. PIE finds payoffs by considering solutions of these constraints and their variants via three heuristics. First, we approximately compute a centroid of the resulting polytope of the constraints. Second, we use a soft constraint approach that allows violation of constraints by penalizing violations in the objective function. Third, we develop a novel approach to payoff inference based on support vector machines (SVMs). Unlike past work on payoff inference, PIE has the following advantages. PIE supports reasoning about multiplayer games, not just one or two players, it can use short histories, not long ones which may not be available in many real-world situations, it does not require all players to be fully rational, and it is one to two orders of magnitude more scalable than past work. We run experiments on a synthetic data set where we generate payoff functions for the players and see how well our algorithms can learn them, a real-world coarse-grained counter-terrorism data set about a set of different terrorist groups, and a real-world fine-grained data set about a specific terrorist group. As the ground truth about payoffs for the terrorist groups cannot be tested directly, we test PIE by using the payoffs to make predictions about the actions of the groups and corresponding governments (even though this is not the purpose of this article). We show that compared with recent work on payoff inference, PIE has both higher accuracy and much shorter runtime.

When a new vulnerability is discovered, a Common Vulnerability and Exposure (CVE) number is publicly assigned to it. The vulnerability is then analyzed by the US National Institute of Standards and Technology (NIST) whose Common Vulnerability Scoring System (CVSS) evaluates a severity score that ranges from 0 to 10 for the vulnerability. On average, NIST takes 132.7 days for this - but early knowledge of the CVSS score is critical for enterprise security managers to take defensive actions (e.g. patch prioritization). We present VASE (Vulnerability Analysis and Scoring Engine) that uses Twitter discussions about CVEs to predict CVSS scores before the official assessments from NIST. In order to leverage the intrinsic correlations between different vulnerabilities, VASE adopts a graph convolutional network (GCN) model in which nodes correspond to CVEs. In addition, we propose a novel attention-based input embedding method to extract useful latent features for each CVE node. We show on real-world data that VASE obtains a mean absolute error (MAE) of 1.255 for predicting the CVSS score using only three days of Twitter discussion data after the date a vulnerability is first mentioned on Twitter. VASE can provide predictions for the CVSS scores for 37.85% of the CVEs at least one week earlier than the official assessments by NIST.

For both the traditional street-hailing taxi industry and the recently emerged on-line ride-hailing, it has been a major challenge to improve the ride-hailing marketplace efficiency due to spatio-temporal imbalance between the supply and demand, among other factors. Despite the numerous approaches to improve marketplace efficiency using pricing and dispatch strategies, they usually optimize pricing or dispatch separately. In this paper, we show that these two processes are in fact intrinsically interrelated. Motivated by this observation, we make an attempt to simultaneously optimize pricing and dispatch strategies. However, such a joint optimization is extremely challenging due to the inherent huge scale and lack of a uniform model of the problem. To handle the high complexity brought by the new problem, we propose InBEDE (Integrating contextual Bandit with tEmporal DiffErence learning), a learning framework where pricing strategies are learned via a contextual bandit algorithm, and the dispatch strategies are optimized with the help of temporal difference learning. The two learning components proceed in a mutual bootstrapping manner, in the sense that the policy evaluations of the two components are inter-dependent. Evaluated with real-world datasets of two Chinese cities from Didi Chuxing, an online ride-hailing platform, we show that the market efficiency of the ride-hailing platform can be significantly improved using InBEDE.

In many cases, an organization wishes to release some data, but is restricted in the amount of data to be released due to legal, privacy and other concerns. For instance, the US Census Bureau releases only 1% of its table of records every year, along with statistics about the entire table. However, the machine learning (ML) models trained on the released sub-table are usually sub-optimal. In this paper, our goal is to find a way to augment the sub-table by generating a synthetic table from the released sub-table, under the constraints that the generated synthetic table (i) has similar statistics as the entire table, and (ii) preserves the functional dependencies of the released sub-table. We propose a novel generative adversarial network framework called ITS-GAN, where both the generator and the discriminator are specifically designed to satisfy these two constraints. By evaluating the augmentation performance of ITS-GAN on two representative datasets, the US Census Bureau data and US Bureau of Transportation Statistics (BTS) data, we show that ITS-GAN yields high quality classification results, and significantly outperforms various state-of-the-art data augmentation approaches.

Knowing if/when a cyber-vulnerability will be exploited and how severe the vulnerability is can help enterprise security officers (ESOs) come up with appropriate patching schedules. Today, this ability is severely compromised: our study of data from Mitre and NIST shows that on average there is a 132 day gap between the announcement of a vulnerability by Mitre and the time NIST provides an analysis with severity score estimates and 8 important severity attributes. Many attacks happen during this very 132-day window. We present Vulnerability Exploit Scoring \& Timing (VEST), a system for (early) prediction and visualization of if/when a vulnerability will be exploited, and its estimated severity attributes and score.

Over the past decades, Electronic Toll Collection (ETC) systems have been proved the capability of alleviating traffic congestion in urban areas. Dynamic Electronic Toll Collection (DETC) was recently proposed to further improve the efficiency of ETC, where tolls are dynamically set based on traffic dynamics. However, computing the optimal DETC scheme is computationally difficult and existing approaches are limited to small scale or partial road networks, which significantly restricts the adoption of DETC. To this end, we propose a novel multi-agent reinforcement learning (RL) approach for DETC. We make several key contributions: i) an enhancement over the state-of-the-art RL-based method with a deep neural network representation of the policy and value functions and a temporal difference learning framework to accelerate the update of target values, ii) a novel edge-based graph convolutional neural network (eGCN) to extract the spatio-temporal correlations of the road network state features, iii) a novel cooperative multi-agent reinforcement learning (MARL) which divides the whole road network into partitions according to their geographic and economic characteristics and trains a tolling agent for each partition. Experimental results show that our approach can scale up to realistic-sized problems with robust performance and significantly outperform the state-of-the-art method.

When a new cyber-vulnerability is detected, a Common Vulnerability and Exposure (CVE) number is attached to it. Malicious "exploits'' may use these vulnerabilities to carry out attacks. Unlike works which study if a CVE will be used in an exploit, we study the problem of predicting when an exploit is first seen. This is an important question for system administrators as they need to devote scarce resources to take corrective action when a new vulnerability emerges. Moreover, past works assume that CVSS scores (released by NIST) are available for predictions, but we show on average that 49% of real world exploits occur before CVSS scores are published. This means that past works, which use CVSS scores, miss almost half of the exploits. In this paper, we propose a novel framework to predict when a vulnerability will be exploited via Twitter discussion, without using CVSS score information. We introduce the unique concept of a family of CVE-Author-Tweet (CAT) graphs and build a novel set of features based on such graphs. We define recurrence relations capturing "hotness" of tweets, "expertise" of Twitter users on CVEs, and "availability" of information about CVEs, and prove that we can solve these recurrences via a fix point algorithm. Our second innovation adopts Hawkes processes to estimate the number of tweets/retweets related to the CVEs. Using the above two sets of novel features, we propose two ensemble forecast models FEEU (for classification) and FRET (for regression) to predict when a CVE will be exploited. Compared with natural adaptations of past works (which predict if an exploit will be used), FEEU increases F1 score by 25.1%, while FRET decreases MAE by 37.2%.

To alleviate traffic congestion in urban areas, electronic toll collection (ETC) systems are deployed all over the world. Despite the merits, tolls are usually pre-determined and fixed from day to day, which fail to consider traffic dynamics and thus have limited regulation effect when traffic conditions are abnormal. In this paper, we propose a novel dynamic ETC (DyETC) scheme which adjusts tolls to traffic conditions in realtime. The DyETC problem is formulated as a Markov decision process (MDP), the solution of which is very challenging due to its 1) multi-dimensional state space, 2) multi-dimensional, continuous and bounded action space, and 3) time-dependent state and action values. Due to the complexity of the formulated MDP, existing methods cannot be applied to our problem. Therefore, we develop a novel algorithm, PG-beta, which makes three improvements to traditional policy gradient method by proposing 1) time-dependent value and policy functions, 2) Beta distribution policy function and 3) state abstraction. Experimental results show that, compared with existing ETC schemes, DyETC increases traffic volume by around 8%, and reduces travel time by around 14:6% during rush hour. Considering the total traffic volume in a traffic network, this contributes to a substantial increase to social welfare.

It has been an open challenge for self-interested agents to make optimal sequential decisions in complex multiagent systems, where agents might achieve higher utility via collaboration. The Microsoft Malmo Collaborative AI Challenge (MCAC), which is designed to encourage research relating to various problems in Collaborative AI, takes the form of a Minecraft mini-game where players might work together to catch a pig or deviate from cooperation, for pursuing high scores to win the challenge. Various characteristics, such as complex interactions among agents, uncertainties, sequential decision making and limited learning trials all make it extremely challenging to find effective strategies. We present HogRider---the champion agent of MCAC in 2017 out of 81 teams from 26 countries. One key innovation of HogRider is a generalized agent type hypothesis framework to identify the behavior model of the other agents, which is demonstrated to be robust to observation uncertainty. On top of that, a second key innovation is a novel Q-learning approach to learn effective policies against each type of the collaborating agents. Various ideas are proposed to adapt traditional Q-learning to handle complexities in the challenge, including state-action abstraction to reduce problem scale, a warm start approach using human reasoning for addressing limited learning trials, and an active greedy strategy to balance exploitation-exploration. Challenge results show that HogRider outperforms all the other teams by a significant edge, in terms of both optimality and stability.

In cloud computing, a private (secondary) cloud can: 1) outsource workload to public (primary) clouds via vertical federation or 2) share resources with other secondary clouds through horizontal federation to enhance its service quality. While there have been attempts to establish a joint vertical and horizontal cloud federation (VHCF), little is known regarding the economic aspects (e.g., what stable cooperation pattern will form, will it improve efficiency) of such a complex cloud network, where secondary clouds are self-interested. To fill the gap, we analyze the interrelated workload factoring and federation formation among secondary clouds, while providing scalable algorithms to assist them to optimally select partners and outsource workload. We use a game theoretic approach to model the federation formation of clouds as a coalition game with externalities. We adopt a pessimistic core to characterize the cooperation stability and formulate its computation as a bilevel optimization problem. The properties of the problem are explored and efficient algorithms are developed to solve it. Experimental results show that the two common practices (no-cooperation and all-in-one federation) are not always stable. The results also show that compared with the two common practices, secondary clouds can decrease service delay penalty by around 11% with the proposed VHCF network.

Although numerous traditional models predict market share and demand along airline routes, the prediction of existing models is not precise enough, and to the best of our knowledge, there is no use of data mining--based forecasting techniques for improving airline profitability. We propose the maximizing airline profits (MAP) architecture designed to help airlines and make two key contributions in airline market share and route demand prediction and prediction-based airline profit optimization. Compared to past methods used to forecast market share and demand along airline routes, we introduce a novel ensemble forecasting (MAP-EF) approach considering two new classes of features: (i) features derived from clusters of similar routes and (ii) features based on equilibrium pricing. We show that MAP-EF achieves much better Pearson correlation coefficients (greater than 0.95 vs. 0.82 for market share, 0.98 vs. 0.77 for demand) and R 2 -values compared to three state-of-the-art works for forecasting market share and demand while showing much lower variance. Using the results of MAP-EF, we develop MAP--bilevel branch and bound (MAP-BBB) and MAP-greedy (MAP-G) algorithms to optimally allocate flight frequencies over multiple routes to maximize an airline’s profit. We also study two extensions of the profit maximization problem considering frequency constraints and long-term profits. Furthermore, we develop algorithms for computing Nash equilibrium frequencies when there are multiple strategic airlines. Experimental results show that airlines can increase profits by a significant margin. All experiments were conducted with data aggregated from four sources: the U.S. Bureau of Transportation Statistics (BTS), the U.S. Bureau of Economic Analysis (BEA), the National Transportation Safety Board (NTSB), and the U.S. Census Bureau (CB).

Though there are numerous traditional models to predict market share and demand along airline routes, the prediction of existing models is not precise enough and, to the best of our knowledge, there is no use of data-mining based forecasting techniques to improve airline profitability. We propose the MAP (Maximizing Airline Profits) architecture designed to help airlines and make two key contributions in airline market share and route demand prediction and prediction-based airline profit optimization. Compared with past methods to forecast market share and demand along airline routes, we introduce a novel Ensemble Forecasting (MAP-EF) approach considering two new classes of features: (i) features derived from clusters of similar routes, and (ii) features based on equilibrium pricing. We show that MAP-EF achieves much better Pearson Correlation Coefficients (over 0.95 vs. 0.82 for market share, 0.98 vs. 0.77 for demand) and R2-values compared with three state-of-the-art works for forecasting market share and demand, while showing much lower variance. Using the results of MAP-EF, we develop MAP-Bilevel Branch and Bound (MAP-BBB) and MAP-Greedy (MAP-G) algorithms to optimally allocate flight frequencies over multiple routes, to maximize an airline's profit. Experimental results show that airlines can increase profits by a significant margin. All experiments were conducted with data aggregated from four sources: US Bureau of Transportation Statistics (BTS), US Bureau of Economic Analysis (BEA), the National Transportation Safety Board (NTSB), and the US Census Bureau (CB).