Results 31 to 40 of about 15,576,316 (273)
Disentangled Information Bottleneck
Proceedings of the AAAI Conference on Artificial Intelligence, 2021 The information bottleneck (IB) method is a technique for extracting information that is relevant for predicting the target random variable from the source random variable, which is typically implemented by optimizing the IB Lagrangian that balances the compression and prediction terms.
Ziqi Pan +3 more
openaire +3 more sources
State predictive information bottleneck [PDF]
The Journal of Chemical Physics, 2021 The ability to make sense of the massive amounts of high-dimensional data generated from molecular dynamics simulations is heavily dependent on the knowledge of a low-dimensional manifold (parameterized by a reaction coordinate or RC) that typically distinguishes between relevant metastable states, and which captures the relevant slow dynamics of ...
Dedi Wang, Pratyush Tiwary
openaire +3 more sources
Recognizable Information Bottleneck
Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023 Information Bottlenecks (IBs) learn representations that generalize to unseen data by information compression. However, existing IBs are practically unable to guarantee generalization in real-world scenarios due to the vacuous generalization bound. The recent PAC-Bayes IB uses information complexity instead of information compression to establish a ...
Yilin Lyu +6 more
openaire +2 more sources
Multivariate Information Bottleneck [PDF]
Neural Computation, 2006 The information bottleneck (IB) method is an unsupervised model independent data organization technique. Given a joint distribution, p(X, Y), this method constructs a new variable, T, that extracts partitions, or clusters, over the values of X that are informative about Y. Algorithms that are motivated by the IB method have already been applied to text
Noam Slonim +2 more
openaire +3 more sources
Embo: a Python package for empirical data analysis using the Information Bottleneck
Journal of Open Research Software, 2021 We present 'embo', a Python package to analyze empirical data using the Information Bottleneck (IB) method and its variants, such as the Deterministic Information Bottleneck (DIB).
Eugenio Piasini +3 more
doaj +1 more source
The Information Bottleneck's Ordinary Differential Equation: First-Order Root Tracking for the Information Bottleneck. [PDF]
Entropy (Basel), 2023 The Information Bottleneck (IB) is a method of lossy compression of relevant information. Its rate-distortion (RD) curve describes the fundamental tradeoff between input compression and the preservation of relevant information embedded in the input ...
Agmon S.
europepmc +2 more sources
Information Bottleneck Analysis by a Conditional Mutual Information Bound
Entropy, 2021 Task-nuisance decomposition describes why the information bottleneck loss I(z;x)−βI(z;y) is a suitable objective for supervised learning. The true category y is predicted for input x using latent variables z.
Taro Tezuka, Shizuma Namekawa
doaj +1 more source
Collaborative Information Bottleneck [PDF]
IEEE Transactions on Information Theory, 2019 Submitted to IEEE Transactions on Information Theory (revised, 29, 7 figures)
Matías Vera +2 more
openaire +4 more sources
CoRR, 2020 Representation learning of graph-structured data is challenging because both graph structure and node features carry important information. Graph Neural Networks (GNNs) provide an expressive way to fuse information from network structure and node features. However, GNNs are prone to adversarial attacks.
Tailin Wu +3 more
openaire +3 more sources
Heterogeneous Graph Information Bottleneck [PDF]
Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, 2021 Most attempts on extending Graph Neural Networks (GNNs) to Heterogeneous Information Networks (HINs) implicitly take the direct assumption that the multiple homogeneous attributed networks induced by different meta-paths are complementary. The doubts about the hypothesis of complementary motivate an alternative assumption of consensus. That is, the
Liang Yang 0002 +7 more
openaire +2 more sources