Towards a Consistent Interpretation of AIOps Models

Artificial Intelligence for IT Operations (AIOps) has been adopted in organizations in various tasks, including interpreting models to identify indicators of service failures. To avoid misleading practitioners, AIOps model interpretations should be consistent (i.e., different AIOps models on the same task agree with one another on feature importance). However, many AIOps studies violate established practices in the machine learning community when deriving interpretations, such as interpreting models with suboptimal performance, though the impact of such violations on the interpretation consistency has not been studied. In this article, we investigate the consistency of AIOps model interpretation along three dimensions: internal consistency, external consistency, and time consistency. We conduct a case study on two AIOps tasks: predicting Google cluster job failures and Backblaze hard drive failures. We find that the randomness from learners, hyperparameter tuning, and data sampling should be controlled to generate consistent interpretations. AIOps models with AUCs greater than 0.75 yield more consistent interpretation compared to low-performing models. Finally, AIOps models that are constructed with the Sliding Window or Full History approaches have the most consistent interpretation with the trends presented in the entire datasets. Our study provides valuable guidelines for practitioners to derive consistent AIOps model interpretation.

How time-inconsistent preferences influence venture capital exit decisions? A new perspective for grandstanding

Considering that the assumption of time consistency does not adequately reveal the mechanisms of exit decisions of venture capital (VC), this study proposes two kinds of time-inconsistent preferences (i.e., time-flow inconsistency and time-point inconsistency) to advance research in this field. Time-flow inconsistency is in line with the previous time inconsistency literature, while time-point inconsistency is rooted in the VC fund’s finite lifespan. Based on the assumption about the strategies guiding future behaviors, we consider four types of venture capitalists: time-consistent, time-point-inconsistent, naïve, and sophisticated venture capitalists, of which the latter three are time-inconsistent. We derive and compare the exit thresholds of these four types of venture capitalists. The main results include: (1) time-inconsistent preferences accelerate the exits of venture capitalists; (2) the closer the VC funds expiry dates are, the more likely time-inconsistent venture capitalists are to accelerate their exits; and (3) future selves caused by time-flow inconsistency weaken the effect of time-point inconsistency. Our study provides a behavioral explanation for the empirical fact of young VCs’ grandstanding.

Strong Time-Consistency of a Core: An Application to the Pollution Control Problem in Eastern Siberia

In this paper, we study the (strong) time-consistency property of the core for a linear-quadratic differential game of pollution control with nonzero absorption coefficient and real values of the model parameters. The values of parameters are evaluated based on the data for the largest aluminum enterprises of Eastern Siberia region of the Russian Federation for the year 2016. The obtained results are accompanied with illustrations.

Time consistency for scalar multivariate risk measures

In this paper we present results on dynamic multivariate scalar risk measures, which arise in markets with transaction costs and systemic risk. Dual representations of such risk measures are presented. These are then used to obtain the main results of this paper on time consistency; namely, an equivalent recursive formulation of multivariate scalar risk measures to multiportfolio time consistency. We are motivated to study time consistency of multivariate scalar risk measures as the superhedging risk measure in markets with transaction costs (with a single eligible asset) (Jouini and Kallal (1995), Löhne and Rudloff (2014), Roux and Zastawniak (2016)) does not satisfy the usual scalar concept of time consistency. In fact, as demonstrated in (Feinstein and Rudloff (2021)), scalar risk measures with the same scalarization weight at all times would not be time consistent in general. The deduced recursive relation for the scalarizations of multiportfolio time consistent set-valued risk measures provided in this paper requires consideration of the entire family of scalarizations. In this way we develop a direct notion of a “moving scalarization” for scalar time consistency that corroborates recent research on scalarizations of dynamic multi-objective problems (Karnam, Ma and Zhang (2017), Kováčová and Rudloff (2021)).

Routing Optimization with Generalized Consistency Requirements

The consistent vehicle routing problem (ConVRP) aims to design synchronized routes on multiple days to serve a group of customers while minimizing the total travel cost. It stipulates that customers should be visited at roughly the same time (time consistency) by several familiar drivers (driver consistency). This paper generalizes the ConVRP for any level of driver consistency and additionally addresses route consistency, which means that each driver can traverse at most a certain proportion of different arcs of routes on planning days, which guarantees route familiarity. To solve this problem, we develop two set partitioning-based formulations, one based on routes and the other based on schedules. We investigate valid lower bounds on the linear relaxations of both of the formulations that are used to derive a subset of columns (routes and schedules); within the subset are columns of an optimal solution for each formulation. We then solve the reduced problem of either one of the formulations to achieve an optimal solution. Numerical results show that our exact method can effectively solve most of the medium-sized ConVRP instances in the literature and can also solve some newly generated instances involving up to 50 customers. Our exact solutions explore some managerial findings with respect to the adoption of consistency measures in practice. First, maintaining reasonably high levels of consistency requirements does not necessarily always lead to a substantial increase in cost. Second, a high level of time consistency can potentially be guaranteed by adopting a high level of driver consistency. Third, maintaining high levels of time consistency and driver consistency may lead to lower levels of route consistency.

Environmental Regulation of Firms that Experience Dynamic Inconsistency

<p>The recent push for environmental regulation has invigorated the discussion of mechanism design and optimal taxation policy. Recent decades have also seen growing interest in behavioural economics and empirically based theory. In this thesis we take a step towards combining the two by asking how a regulator may correct an externality in situations where they have a time consistency problem. Time inconsistency is one of the notable developments of behavioural economics. It posits that an agent’s decisions do not remain consistent over time, which causes a utility loss if the agent cannot commit themselves to a particular course of action and stick to it. The solution to inconsistency problems is to precommit to a course of action and prevent future deviations from it. However, finding a mechanism to enable such precommitment is often problematic. A regulator who maximises welfare can have a time consistency problem because welfare will depend on the decisions of firm and households who may themselves be inconsistent. That inconsistency then propagates to the regulator’s decision and reduces the level of welfare that the regulator can reach. Alternatively, the regulator’s time consistency problem can be caused by non-stationarity in their time preferences. To reach the firstbest outcome the regulator must not only eliminate the environmental externality: they must also overcome their own time inconsistency problem. This thesis draws from the literature on strategic delegation to construct a taxation game in which the regulator can achieve the first best taxation regime without the need for external precommitment devices. We study a dynamic game where the regulator chooses a tax rate and the regulated monopolist chooses their price. We show that the Markov-perfect equilibrium price path of this game will replicate the first best plan. Our results holds for time inconsistency caused by both jump states and quasihyperbolic discounting.</p>

Environmental Regulation of Firms that Experience Dynamic Inconsistency

<p>The recent push for environmental regulation has invigorated the discussion of mechanism design and optimal taxation policy. Recent decades have also seen growing interest in behavioural economics and empirically based theory. In this thesis we take a step towards combining the two by asking how a regulator may correct an externality in situations where they have a time consistency problem. Time inconsistency is one of the notable developments of behavioural economics. It posits that an agent’s decisions do not remain consistent over time, which causes a utility loss if the agent cannot commit themselves to a particular course of action and stick to it. The solution to inconsistency problems is to precommit to a course of action and prevent future deviations from it. However, finding a mechanism to enable such precommitment is often problematic. A regulator who maximises welfare can have a time consistency problem because welfare will depend on the decisions of firm and households who may themselves be inconsistent. That inconsistency then propagates to the regulator’s decision and reduces the level of welfare that the regulator can reach. Alternatively, the regulator’s time consistency problem can be caused by non-stationarity in their time preferences. To reach the firstbest outcome the regulator must not only eliminate the environmental externality: they must also overcome their own time inconsistency problem. This thesis draws from the literature on strategic delegation to construct a taxation game in which the regulator can achieve the first best taxation regime without the need for external precommitment devices. We study a dynamic game where the regulator chooses a tax rate and the regulated monopolist chooses their price. We show that the Markov-perfect equilibrium price path of this game will replicate the first best plan. Our results holds for time inconsistency caused by both jump states and quasihyperbolic discounting.</p>

Lane and Road Marker Semantic Video Segmentation Using Mask Cropping and Optical Flow Estimation

Lane and road marker segmentation is crucial in autonomous driving, and many related methods have been proposed in this field. However, most of them are based on single-frame prediction, which causes unstable results between frames. Some semantic multi-frame segmentation methods produce error accumulation and are not fast enough. Therefore, we propose a deep learning algorithm that takes into account the continuity information of adjacent image frames, including image sequence processing and an end-to-end trainable multi-input single-output network to jointly process the segmentation of lanes and road markers. In order to emphasize the location of the target with high probability in the adjacent frames and to refine the segmentation result of the current frame, we explicitly consider the time consistency between frames, expand the segmentation region of the previous frame, and use the optical flow of the adjacent frames to reverse the past prediction, then use it as an additional input of the network in training and reasoning, thereby improving the network’s attention to the target area of the past frame. We segmented lanes and road markers on the Baidu Apolloscape lanemark segmentation dataset and CULane dataset, and present benchmarks for different networks. The experimental results show that this method accelerates the segmentation speed of video lanes and road markers by 2.5 times, increases accuracy by 1.4%, and reduces temporal consistency by only 2.2% at most.

Dynamic Bargaining and Time-Consistency in Linear-State and Homogeneous Linear-Quadratic Cooperative Differential Games

Three different solution concepts are reviewed and computed for linear-state and homogeneous linear-quadratic cooperative differential games with asymmetric players. Discount rates can be nonconstant and/or different. Special attention is paid to the issues of time-consistency, agreeability and subgame-perfectness, both from the viewpoint of sustainability of cooperation and from the credibility of the announced equilibrium strategies.