TraceVis: Towards Visualization for Deep Statistical Model Checking

Gros, Timo P.; Gross, David C.; Gumhold, Stefan; Hoffmann, Jörg; Klauck, Michaela; Steinmetz, Marcel

doi:10.1007/978-3-030-83723-5_3

Cited by 9 publications

(6 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our Racetrack case study makes it easy to produce "heat maps", as a meaningful way to represent a partitioned perspective on the state space and sampling one member state from each set as a representative. With the TraceVis tool, we also showed how visualization techniques in 3D can help to get even more insights from the DSMC results and to display more information than in the simple heat maps [26,28]. We believe that such a representative analysis makes sense (e.g., to provide an overview for human users) in many application scenarios.…”

Section: Discussionmentioning

confidence: 99%

“…There are already works building up on DSMC giving evidence for the potential impact of the approach. The information delivered by DSMC has already been used to improve reinforcement learning strategies [32] and for the design of policy-analysis tools in synergy with interactive visualization techniques [26,28]. The most important work based on DSMC is MoGym [29], the integrated toolbox enabling the training and verification of machine-learned decisionmaking agents based on formal models, which bridges the reinforcement learning community to formal methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Analyzing neural network behavior through deep statistical model checking

Gros

Hermanns

Hoffmann

et al. 2022

Int J Softw Tools Technol Transfer

Self Cite

View full text Add to dashboard Cite

Neural networks (NN) are taking over ever more decisions thus far taken by humans, even though verifiable system-level guarantees are far out of reach. Neither is the verification technology available, nor is it even understood what a formal, meaningful, extensible, and scalable testbed might look like for such a technology. The present paper is an attempt to improve on both the above aspects. We present a family of formal models that contain basic features of automated decision-making contexts and which can be extended with further orthogonal features, ultimately encompassing the scope of autonomous driving. Due to the possibility to model random noise in the decision actuation, each model instance induces a Markov decision process (MDP) as verification object. The NN in this context has the duty to actuate (near-optimal) decisions. From the verification perspective, the externally learnt NN serves as a determinizer of the MDP, the result being a Markov chain which as such is amenable to statistical model checking. The combination of an MDP and an NN encoding the action policy is central to what we call “deep statistical model checking” (DSMC). While being a straightforward extension of statistical model checking, it enables to gain deep insight into questions like “how high is the NN-induced safety risk?”, “how good is the NN compared to the optimal policy?” (obtained by model checking the MDP), or “does further training improve the NN?”. We report on an implementation of DSMC inside the ModestToolset in combination with externally learnt NNs, demonstrating the potential of DSMC on various instances of the model family, and illustrating its scalability as a function of instance size as well as other factors like the degree of NN training.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Analyzing neural network behavior through deep statistical model checking

Gros

Hermanns

Hoffmann

et al. 2022

Int J Softw Tools Technol Transfer

Self Cite

View full text Add to dashboard Cite

show abstract

“…Deep RL with DSMC Specifics. Usually, learning NN is done on GPUs [43][44][45][46], but for a reasonable runtime comparison, we used a CPU infrastructure here. In addition, the random start setup [44]-during learning, the agent starts randomly from one of the free road cells instead of always from the same start cell-leads to significantly better learning performance.…”

Section: Methodsmentioning

confidence: 99%

“…But since the other methods we compare to can only start from a fixed cell, we used the normal start setup during learning for this paper, where the agent also starts its exploration runs from a single start position always. The NN we trained have an input layer of 15 neurons, two hidden layers of 64 neurons each and an output layer of 9 neurons encoding the nine possible acceleration values, as done in other case studies on Racetrack [43,44,46]. We start with the barto-small track shown in Fig.…”

Section: Methodsmentioning

confidence: 99%

“…The DSMC extension of modes is also integrated in DSMC evaluation stages [46], where DSMC is applied during deep RL to determine state space regions with weak performance to concentrate on during the learning process. To visualise the SMC results of modes when executing DSMC on Racetrack benchmarks, the tool TraceVis has been implemented [43]. It takes the traces generated by modes as input, visualises and clusters them, and provides information on the goal probability when starting on a predefined position.…”

Section: Deep Statistical Model Checkingmentioning

confidence: 99%

See 1 more Smart Citation

The Modest State of Learning, Sampling, and Verifying Strategies

Hartmanns

Klauck

2022

Leveraging Applications of Formal Methods, Verification and Validation. Adaptation and Learning

Self Cite

View full text Add to dashboard Cite

Optimal decision-making under stochastic uncertainty is a core problem tackled in artificial intelligence/machine learning (AI), planning, and verification. Planning and AI methods aim to find good or optimal strategies to maximise rewards or the probability of reaching a goal. Verification approaches focus on calculating the probability or reward, obtaining the strategy as a side effect. In this paper, we connect three strands of work on obtaining strategies implemented in the context of the Modest Toolset: statistical model checking with either lightweight scheduler sampling or deep learning, and probabilistic model checking. We compare their different goals and abilities, and show newly extended experiments on Racetrack benchmarks that highlight the tradeoffs between the methods. We conclude with an outlook on improving the existing approaches and on generalisations to continuous models, and emphasise the need for further tool development to integrate methods that find, evaluate, compare, and explain strategies.

show abstract

Momba: JANI Meets Python

Köhl

Klauck

Hermanns

2021

Tools and Algorithms for the Construction and Analysis of Systems

View full text Add to dashboard Cite

JANI-model [6] is a model interchange format for networks of interacting automata. It is well-entrenched in the quantitative model checking community and allows modeling a variety of systems involving concurrency, probabilistic and real-time aspects, as well as continuous dynamics. Python is a general purpose programming language preferred by many for its ease of use and vast ecosystem. In this paper, we present Momba, a flexible Python framework for dealing with formal models centered around the JANI-model format and formalism. Momba strives to deliver an integrated and intuitive experience for experimenting with formal models making them accessible to a broader audience. To this end, it provides a pythonic interface for model construction, validation, and analysis. Here, we demonstrate these capabilities.

show abstract

TraceVis: Towards Visualization for Deep Statistical Model Checking

Cited by 9 publications

References 34 publications

Analyzing neural network behavior through deep statistical model checking

Analyzing neural network behavior through deep statistical model checking

The Modest State of Learning, Sampling, and Verifying Strategies

Momba: JANI Meets Python

Contact Info

Product

Resources

About