Algorithms for Verifying Deep Neural Networks

Liu, Changliu; Arnon, Tomer; Lazarus, Christopher; Strong, Christopher A.; Barrett, Clark; Kochenderfer, Mykel J.

doi:10.1561/9781680837872

Cited by 22 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be used to detect learning difficulties that students may encounter. By analyzing students' learning behavior patterns, response times, error patterns, etc., it can predict that students may encounter difficulties in specific topics or concepts, and provide corresponding help and support as early as possible [12][13]. Behavior prediction can also be used to plan students' personalized learning paths.…”

Section: Learning Behavior Predictionmentioning

confidence: 99%

Learning Behavior Analysis and Prediction of Teaching System Based on Neural Network Algorithm

Dong

2024

Proceedings of the First International Conference on Science, Engineering and Technology Practices for Sustainable Development,

View full text Add to dashboard Cite

The current teaching system relies on students' behavior and interaction data in the system. This article proposes a teaching system based on recurrent neural networks to solve the above problems. This article collects learners' behavioral data and applies appropriate neural network models for training and analysis. Teaching systems based on neural networks can accurately capture learners' behavioral patterns and trends, and gain an in-depth understanding of their learning characteristics and behavioral preferences. Experimental results show that the accuracy of this system is 83%-94%. By analyzing learning behavior, we are able to predict learners' future learning needs and performance and provide personalized learning support. This personalized recommendation and suggestion function can help educators provide accurate learning resources and guidance, and improve learners' learning effects and satisfaction. The research also points out the need to further improve the accuracy and interpretability of the algorithm and apply these research results to actual educational environments to achieve better teaching results. This study provides useful guidance and research basis for teaching systems based on neural network algorithms.

show abstract

Section: Learning Behavior Predictionmentioning

confidence: 99%

Learning Behavior Analysis and Prediction of Teaching System Based on Neural Network Algorithm

Dong

2024

Proceedings of the First International Conference on Science, Engineering and Technology Practices for Sustainable Development,

View full text Add to dashboard Cite

show abstract

“…Reachability is often studied in conjunction with local robustness properties, i.e., whether for a given input, e.g., an image, small alterations of this input can cause output variation, e.g., a different classification. The present state of the art (Liu et al 2019) includes several ways of formulating this problem (see related work below); however, no method scales to the analysis of the neural networks presently used in industrial strength applications, including autonomous vehicles. Therefore, it remains of considerable importance to develop more scalable approaches.…”

Section: Introductionmentioning

confidence: 99%

Efficient Verification of ReLU-Based Neural Networks via Dependency Analysis

Botoeva

Kouvaros

Kronqvist

et al. 2020

AAAI

104

View full text Add to dashboard Cite

We introduce an efficient method for the verification of ReLU-based feed-forward neural networks. We derive an automated procedure that exploits dependency relations between the ReLU nodes, thereby pruning the search tree that needs to be considered by MILP-based formulations of the verification problem. We augment the resulting algorithm with methods for input domain splitting and symbolic interval propagation. We present Venus, the resulting verification toolkit, and evaluate it on the ACAS collision avoidance networks and models trained on the MNIST and CIFAR-10 datasets. The experimental results obtained indicate considerable gains over the present state-of-the-art tools.

show abstract

“…Solutions to the problem of verification of neural networks [8] have developed rapidly, appearing in the context of the need to build a criterion for neural networks training. Currently, a series of approaches are used to solve the verification problem [9]. However, it is not resolved in general, and the methods do not involve any analysis of the specifics of data distribution within the input and output data sets.…”

Section: Introductionmentioning

confidence: 99%

Artificial Neural Network Training Criterion Formulation Using Error Continuous Domain

Hu¹,

Ivashchenko²,

Lyushenko³

et al. 2021

IJMECS

View full text Add to dashboard Cite

One of the trends in information technologies is implementing neural networks in modern software packages [1]. The fact that neural networks cannot be directly programmed (but trained) is their distinctive feature. In this regard, the urgent task is to ensure sufficient speed and quality of neural network training procedures. The process of neural network training can differ significantly depending on the problem. There are verification methods that correspond to the task's constraints; they are used to assess the training results. Verification methods provide an estimate of the entire cardinal set of examples but do not allow to estimate which subset of those causes a significant error. This fact leads to neural networks' failure to perform with the given set of hyperparameters, making training a new one time-consuming.On the other hand, existing empirical assessment methods of neural networks training use discrete sets of examples. With this approach, it is impossible to say that the network is suitable for classification on the whole cardinal set of examples.This paper proposes a criterion for assessing the quality of classification results. The criterion is formed by describing the training states of the neural network. Each state is specified by the correspondence of the set of errors to the function range representing a cardinal set of test examples. The criterion usage allows tracking the network's classification defects and marking them as safe or unsafe. As a result, it is possible to formally assess how the training and validation data sets must be altered to improve the network's performance, while existing verification methods do not provide any information on which part of the dataset causes the network to underperform.

show abstract

Algorithms for Verifying Deep Neural Networks

Cited by 22 publications

References 0 publications

Learning Behavior Analysis and Prediction of Teaching System Based on Neural Network Algorithm

Learning Behavior Analysis and Prediction of Teaching System Based on Neural Network Algorithm

Efficient Verification of ReLU-Based Neural Networks via Dependency Analysis

Artificial Neural Network Training Criterion Formulation Using Error Continuous Domain

Contact Info

Product

Resources

About