A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL. The file type is application/pdf
.
Filters
Runtime Monitoring Neuron Activation Patterns [article]
2018 arXiv pre-print
We propose runtime neuron activation pattern monitoring - after the standard training process, one creates a monitor by feeding the training data to the network again in order to store the neuron activation ... Our experiments show that, by adjusting the similarity-threshold for activation patterns, the monitors can report a significant portion of misclassfications to be not supported by training with a small ... High-level workflow on runtime monitoring neuron activation patterns. ...
arXiv:1809.06573v2 fatcat:jfpktlyjfrg2tomuoybpq3d47m
Provably-Robust Runtime Monitoring of Neuron Activation Patterns [article]
2021 arXiv pre-print
The provable robustness is further generalized to cases where monitoring a single neuron can use more than one bit, implying that one can record activation patterns with a fine-grained decision on the ... While recent results in monitoring DNN activation patterns provide a sound guarantee due to building an abstraction out of the training data set, reducing false positives due to slight input perturbation ... ROBUST NEURON ACTIVATION PATTERN MONITORS A. DNN and Runtime Monitors A feed-forward DNN model consists of an input layer, multiple hidden layers and an output layer. ...
arXiv:2011.11959v2 fatcat:vvg3stuwwfgsldkfmdtk6dvaz4
Security For System-On-Chip (SoC) Using Neural Networks [article]
2022 arXiv pre-print
Malicious activities generate abnormal traffic patterns which affect the operation of the system and its performance which cannot be afforded in a computation hungry world. ... Spiking Neural Networks (SNN), Runtime Neural Architecture (RTNA) are some of the neural networks which prevent SoCs from attacks. Finally, the development trends in SoC security are also highlighted. ... At runtime, when a set of inputs is fed to the SOC architecture, RTNA is activated. ...
arXiv:2108.13307v2 fatcat:xlpzd3iyrnhnnbjkjki66xttvq
Towards safety monitoring of ML-based perception tasks of autonomous systems
2020 2020 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW)
We also highlight the main issues regarding monitoring ML in safety-critical environments. ... Therefore, a fault tolerance mechanism, such as a safety monitor (SM), should be applied to guarantee the property correctness of these systems. ... After the standard training process, a runtime monitor is created by feeding the training data to the network again in order to store the neuron on-off activation patterns using binary decision diagrams ...
doi:10.1109/issrew51248.2020.00052 fatcat:exz2ngbkwfcr3aftndthbbokie
Runtime Monitoring for Out-of-Distribution Detection in Object Detection Neural Networks [article]
2022 arXiv pre-print
Runtime monitoring provides a more realistic and applicable alternative to verification in the setting of real neural networks used in industry. ... We extend a runtime-monitoring approach previously proposed for classification networks to perception systems capable of identification and localization of multiple objects. ... The idea was based on Gaussian monitoring of the neuron activation patterns. ...
arXiv:2212.07773v1 fatcat:s4bud5px7raqpmq3ujb3fbpl7q
Neural-symbolic monitoring and adaptation
2015 2015 International Joint Conference on Neural Networks (IJCNN)
Runtime monitors check the execution of a system under scrutiny against a set of formal specifications describing a prescribed behaviour. ... The two core properties for monitoring systems are scalability and adaptability. ... C2 the input potential of an output neuron o can only exceed its threshold, activating o, when at least one hidden neuron h that is connected to o is activated. ...
doi:10.1109/ijcnn.2015.7280713 dblp:conf/ijcnn/PerottiGB15 fatcat:ntwp46d2zrfmpps63qc5abcesy
Customizable Reference Runtime Monitoring of Neural Networks using Resolution Boxes [article]
2021 arXiv pre-print
Runtime monitoring techniques on the neuron activation pattern can be used to detect such inputs. We present an approach for monitoring classification systems via data abstraction. ... monitored layers. ... The verdict of a monitor on a new input is based on the membership test of the induced neuron activation pattern in a pre-established sound over-approximation of neuron activation patterns recorded from ...
arXiv:2104.14435v2 fatcat:tjeknqodo5dorcfjwg3gi6sbza
nn-dependability-kit: Engineering Neural Networks for Safety-Critical Autonomous Driving Systems [article]
2019 arXiv pre-print
monitoring for reasoning whether a decision of a neural network in operation is supported by prior similarities in the training data. ... metrics for indicating sufficient elimination of uncertainties in the product life cycle, (b) formal reasoning engine for ensuring that the generalization does not lead to undesired behaviors, and (c) runtime ... monitoring (Sn10) [7] by first using binary decision diagrams (BDD) [3] to record binarized neuron activation patterns in training time, followed by checking if an activation pattern produced during ...
arXiv:1811.06746v2 fatcat:nyfqvxi3rvga5fkseawou6r6b4
Scalable Process Monitoring through Rules and Neural Networks
2014 International Symposium on Data-Driven Process Discovery and Analysis
In this paper we introduce RuleRunner, a Runtime Verification system for monitoring LTL properties over finite traces. ... By exploiting results from the Neural-Symbolic Integration area, a RuleRunner monitor can be encoded in a recurrent neural network. ... In our system, observations (events) are presented to the monitor as the activation of labelled input neurons. ...
dblp:conf/simpda/PerottiBG14 fatcat:gssofqjpwjdfvnvphtyqmbkwni
Acceleration of the SPADE Method Using a Custom-Tailored FP-Growth Implementation
2021 Frontiers in Neuroinformatics
The SPADE (spatio-temporal Spike PAttern Detection and Evaluation) method was developed to find reoccurring spatio-temporal patterns in neuronal spike activity (parallel spike trains). ... Based on a realistic benchmark data set, we identified that the combination of pattern mining (using the FP-Growth algorithm) and the result filtering account for 85–90% of the method's total runtime. ... With the rapid advancement of electrophysiological recording techniques in the recent decades, scientists are now able to monitor the spiking activity of individual nerve cells in large neuronal populations ...
doi:10.3389/fninf.2021.723406 pmid:34603002 pmcid:PMC8483730 fatcat:x2uw2jmii5blrdf5qh6n7chl5y
Architecting Dependable Learning-enabled Autonomous Systems: A Survey [article]
2019 arXiv pre-print
We consider three technology pillars for architecting dependable autonomy, namely diverse redundancy, information fusion, and runtime monitoring. ... [13] uses a monitor to store the neuron on-off activation patterns of training data as a set using binary decision diagrams (BDD) [10] . ... In operation, the monitor rejects the decision made by the network if the computed activation pattern is not contained in the BDD. 2. Monitor the distribution of predicted classes. ...
arXiv:1902.10590v1 fatcat:4unvep4bwjhk7o4yqjbrhyxqg4
Out-Of-Distribution Detection for Deep Neural Networks with Isolation Forest and Local Outlier Factor
2021 IEEE Access
This is not surprising, since the total number of possible activation patterns grows exponentially with the number of monitored neurons. ... Assuming the Rectified Linear Unit (ReLU) activation function is used, a Boolean variable is used to encode the output of each neuron being monitored, and is assigned value of 1 if the neuron is active ...
doi:10.1109/access.2021.3108451 fatcat:i64u4ocx3vgnlkf7xt7xgomzvu
Outside the Box: Abstraction-Based Monitoring of Neural Networks [article]
2020 arXiv pre-print
We propose a framework to monitor a neural network by observing the hidden layers. ... At runtime, inputs not falling into any of the categories learned during training cannot be classified correctly by the neural network. ... The underlying assumption is that the neurons at the watched layers exhibit a pattern typical for inputs of the same class. The monitor is trained to recognize these patterns. ...
arXiv:1911.09032v3 fatcat:46sd6nourzagbmuzgzvsir7xru
User Activity Recognition in Smart Homes Using Pattern Clustering Applied to Temporal ANN Algorithm
2015 Sensors
The activity recognition is usually done through two steps: activity pattern clustering and activity type decision. ... This paper tries to find the best combination of a pattern clustering method and an activity decision algorithm among various existing works. ... Activities daily living (ADL) of users is monitored and the general activity patterns are modeled according to the user position in his/her environment. ...
doi:10.3390/s150511953 pmid:26007738 pmcid:PMC4481973 fatcat:gruuioehrjehzlqntyyugnp7um
ANN Predicted Apps-Usage Aware Linux Scheduler for Asymmetrical Multi Cluster SoC
2016 Journal of Software
The scheduler monitors multiple system performance metrics at runtime, predicts power dissipation and future workload with an ANN computation block. ... This research introduces an improved Linux scheduler that models and manages several power dissipation problems based on user application usage pattern identified in mobile computing platform. ... The input layer has a number of neurons equivalent to the system metrics monitored by the thread_monitor() and the same amount of neurons in the output layer, the number of neurons in the hidden layer ...
doi:10.17706/jsw.11.7.623-630 fatcat:yl26ro22krhk7bqlhgd7nr5xmm
« Previous Showing results 1 — 15 out of 3,658 results