Opening the black box of neural nets: case studies in stop/top discrimination
Explore this paper's citation graph
Summary
This work introduces techniques for exploring the functionality of a neural network and extracting simple, human-readable approximations to its performance by performing gradient ascent on the input space of the network and producing large populations of artificial events which strongly excite a given classifier.
- Type
- preprint
- Published
- 2018-04-24
- Cited by
- 23
- References
- 271
- Access
- Open access
- OpenAlex
- https://openalex.org/W2798285863
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:55754281
Keywords
Artificial neural network, Classifier (UML), Artificial intelligence, Computer science, Physics
References
- Radiative decays of massive neutrinos
- Model Independent Direct Detection Analyses
- Connection between dark matter abundance and primordial tensor perturbations
- A New Candidate for the Dark Matter
- Decoupled Sectors and Wolf-Rayet Galaxies
- Diurnal modulation signal from dissipative hidden sector dark matter
- Neutron-antineutron oscillations on the lattice
- Probing compressed top squark scenarios at the LHC at 14 TeV
- A Supersymmetric Twin Higgs
- A Detailed Look at the First Results from the Large Underground Xenon (LUX) Dark Matter Experiment
- Observational Physics of Mirror World
- X-ray line from exciting dark matter
- Displaced Dark Matter at Colliders
- Dissipative dark matter and the Andromeda plane of satellites
- Four body decay of the top squark at the upgraded Fermilab Tevatron
- Recovering Particle Masses from Missing Energy Signatures with Displaced Tracks
- Playing tag with ANN: boosted top identification with pattern recognition
- DELPHES, a framework for fast simulation of a generic collider experiment
- Revealing compressed stops using high-momentum recoils
- Understanding Neural Networks Through Deep Visualization
Cited by
- Infrared safety of a neural-net top tagging algorithm
- Learning new physics from a machine
- An operational definition of quark and gluon jets
- Energy flow networks: deep sets for particle jets
- Interpretable deep learning for two-prong jet classification with jet spectra
- Deep-learning jets with uncertainties and more
- Quark-gluon tagging: Machine learning vs detector
- Beyond M_tt: learning to search for a broad tt resonance at the LHC
- CapsNets continuing the convolutional quest
- On the ATLAS top mass measurements and the potential for stealth stop contamination
- Adversarially-trained autoencoders for robust unsupervised new physics searches
- Transparency in Complex Computational Systems
- Learning multivariate new physics
- Per-object systematics using deep-learned calibration
- A guide for deploying Deep Learning in LHC searches: How to achieve optimality and account for uncertainty
- Energy Flow in Particle Collisions
- Mapping machine-learned physics into a human-readable space
- Topological obstructions to autoencoding
- Creating simple, interpretable anomaly detectors for new physics in jet substructure
- Top squark signal significance enhancement by different Machine Learning Algorithms
Related papers
- Explanation of Quantum Entanglement Using Hidden Variables
- Semi-mechanistic modeling of chemical processes with neural networks
- Neural Networks: From Black Box Towards Transparent Box Application To Evapotranspiration Modeling
- On the locality of hidden-variable theories in quantum physics
- MRxaI: Black-Box Explainability for Image Classifiers in a Medical Setting
- Bell's theorem does not eliminate fully causal hidden variables
- Flavor of gluino decay in high-scale supersymmetry
- Lepton flavor violating Higgs boson decays from massive seesaw neutrinos