Word embeddings reveal multiple senses, which can be recovered using sparse coding.
problem Understanding word senses in polysemous words.
method Sparse coding of word embeddings to recover multiple senses.
result Sparse coding can approximate multiple word senses, with each sense associated with a discourse atom.
A new neural network models discourse relations with latent variables.
problem Jointly modeling discourse relations and word sequences.
method Latent variable recurrent neural network for discourse relations.
result Model outperforms state-of-the-art alternatives on discourse classification tasks.
HDSF detects fake news by learning hierarchical discourse-level structures.
problem Detecting fake news articles with minimal annotated corpora.
method Automatically learns and constructs discourse-level structures for fake and real news articles.
result Real and fake news present substantial differences in hierarchical discourse-level structures.
Improved text generation with discourse-aware soft prompts.
problem Efficient fine-tuning methods don't generalize across all generation tasks.
method Hierarchical blocking and attention sparsity on prefix parameters.
result Structured design of prefix parameters yields more coherent generations.
This work speeds up unsupervised sentence learning using paragraph coherence.
problem Training fast unsupervised sentence encoders.
method Discourse-based objective function for neural network training.
result Models trained with this method are faster and perform well.
This study identifies sentence relationships in legal transcripts.
problem Improving understanding of legal case proceedings through sentence relationships.
method Combining machine learning and rule-based approach to classify sentence relationships.
result First study to use discourse relationships for legal court case transcripts.
Adversarial model improves implicit relation classification without explicit connectives.
problem Lack of explicit connectives makes implicit discourse relation classification challenging.
method Feature imitation framework with adversarial training.
result State-of-the-art performance on PDTB benchmark.
The study detects deceptive language in business communication using AI.
problem Deceptive language in business communication.
method Combining classical rhetoric, communication psychology, and linguistic theory with computational textual analysis and transformer models.
result Detection accuracies of over 99% achieved in controlled settings.
Text documents are structured on multiple levels of detail: individual words are related by syntax, but larger units of text are related by discourse structure. Existing language models generally fail to account for discourse structure, but it is crucial if we are to have language models that reward coherence and gener…
New approach to topic modelling with covariates for large text corpora.
problem Complex topic modelling in large text corpora.
method Combining convex NMF with regression for tractable estimation.
result Faster, interpretable, and better inferential justification than generative models.
Research finds correlations between Bitcoin online discourse and price/volume movements.
problem Mapping sentiment to Bitcoin price and volume movements.
method Collected and analyzed data from Bitcointalk.org, news sources, and Reddit communities.
result Weak to moderate correlations between online sentiment and Bitcoin price/volume movements.
Unified theory linking atom-centered and message-passing models for molecular properties.
problem Combining atom-centered and message-passing models for accurate molecular property prediction.
method Generalizing ACDC framework to include multi-centered information, providing a complete linear basis for regression.
result Unified understanding of atom-centered and message-passing models, providing a coherent foundation.
Proposes atomic swaptions for trustless cryptocurrency derivatives.
problem Lack of trustless derivatives for cryptocurrency exchanges.
method Extends atomic swap protocol to include derivatives without oracles.
result Atomic swaptions enable trustless exchange of derivative assets.
Model predicts upcoming discourse referents using linguistic and script knowledge.
problem Predicting upcoming discourse referents based on linguistic knowledge.
method Built a computational model that predicts referents using linguistic knowledge and scripts.
result Script knowledge significantly improves model estimates of human predictions.
Researchers use manifold learning to analyze 4D-STEM data of graphene, revealing atomic structure details.
problem Challenges in processing and interpreting large 4D-STEM datasets, especially for light materials.
method Data-driven manifold learning approaches for visualization and exploration of 4D-STEM datasets.
result Extracted patterns relate to individual atom sites and sublattice structures, effectively discriminating single dopant anomalies.
Machine learning predicts atomization energies accurately from low-fidelity calculations.
problem Predicting accurate atomization energies of organic molecules efficiently.
method Machine learning models trained on low-fidelity B3LYP energies to predict high-fidelity G4MP2 energies.
result Predicted G4MP2 atomization energies within 0.012 eV for molecules with 10-14 heavy atoms.
Deep model captures interactions in online forums.
problem Modeling discursive interactions in online forums.
method Coupled Distributed Topics model with deep architecture and GPU-based inference.
result Model outperforms existing methods in online discourse analysis.
New features for quantum calculations learn N-center Hamiltonian matrix elements.
problem Quantum calculations need features for N-center Hamiltonians, not just atom-centered ones.
method Developed fully equivariant N-center features for machine learning.
result Learned matrix elements of N-center Hamiltonians efficiently.
Bitcoin price prediction models fail to outperform a simple 'today's price' baseline, especially at longer horizons.
problem Lack of robust models that consistently outperform a naive price predictor at various horizons.
method Surveyed peer-reviewed papers, categorized by evaluation methodology, contrasted with social media discourse, and proposed methodological standards.
result No peer-reviewed study has shown robust superiority over the naive baseline across multiple market regimes at short-to-medium horizons.
Neural network learns atomic coordinates from Patterson maps in a simplified case.
problem Training a neural network to infer atomic coordinates from Patterson maps.
method Synthetic data training, centering output maps, removing centrosymmetric inversion, and adding empty space.
result The network can generalize to infer atom positions from Patterson maps not in the training set.
Study compares atom representations in graph neural networks for molecular properties.
problem Incorrect attribution of results in molecular property prediction due to varying atom features.
method Evaluated multiple atom representations on free energy, solubility, and metabolic stability predictions.
result Different atom representations can lead to varying predictive performance in graph neural networks.
We introduce a machine learning model to predict atomization energies of a diverse set of organic molecules, based on nuclear charges and atomic positions only. The problem of solving the molecular Schrödinger equation is mapped onto a non-linear statistical regression problem of reduced complexity. Regression models a…
LC-GAP uses localized Coulomb descriptors for accurate molecular potential predictions.
problem Creating accurate molecular potentials for large molecules.
method Combining localized Coulomb matrix representations with Gaussian approximation potential.
result LC-GAP generates accurate potentials for molecules larger than training data with chemical accuracy.
We prove that the introduction of the class of geometrically atomic bundle maps by Harvey and Lawson in their theory of singular connections is not necessary because an arbitrary map satisfies the conditions of geometric atomicity.
ASLA learns atomic structures using neural networks and reinforcement learning.
problem Designing materials and drugs with desired properties.
method Atomistic structure learning algorithm (ASLA) using a convolutional neural network and reinforcement learning.
result ASLA can predict optimal structural arrangements of atoms for various target properties.
PAM models generate dependent random distributions across groups with overlapping clusters.
problem Generating dependent random distributions across multiple groups.
method Atom skipping in an infinite mixture model.
result Interpretable posterior inference of cluster exclusivity and sharing.
Non-atomic arbitrage exploits price differences on Ethereum and other blockchains, accounting for over 10% of Ethereum's block value.
problem Price differences on decentralized exchanges and centralized exchanges lead to MEV.
method Analyzed non-atomic arbitrage on Ethereum's largest DEXes, identifying its prevalence and impact.
result More than 10% of Ethereum's block value is attributed to non-atomic arbitrage, involving over $132 billion.
Graph neural network predicts protonation energies of oxygen atoms in bio-oil molecules.
problem Predicting protonation energies of oxygen atoms in bio-oil molecules for chemical upgrading.
method Site-specific graph neural network approach using iterative local nonlinear embedding.
result Effective prediction of protonation energies of individual oxygen atoms in bio-oil molecules.
ATOMO reduces communication in distributed learning by sparsifying gradients.
problem Communication overheads in distributed model training.
method ATOMO framework for atomic sparsification of stochastic gradients.
result Sparsifying singular value decomposition can lead to faster distributed training.
Improved chemical predictions through compressed atomic species representations.
problem Intractable chemical space of molecules and materials.
method Introducing elemental modes for compressed representation of atomic species.
result Elemental modes enable improvements in machine learning tasks for chemical predictions.
Unified approach to invariants in equivariant geometry.
problem Constructing invariants in G-equivariant birational geometry. method Combining atom theory and modular symbols.
result Developed new geometric invariants for orbifolds and surfaces.
Study examines active travel in Chicago communities, revealing mixed perceptions.
problem Transport disadvantage and lack of active mobility in underserved communities.
method Focus groups, qualitative discourse analysis, quantitative text-mining (topic modeling, sentiment analysis).
result Residents view active travel as both necessity and symbol of privilege, influenced by local culture.
Improves molecular activity prediction using graph convolutional neural networks considering graph distances.
problem Predicting molecular activity using graph convolutional neural networks with improved distance representation.
method Proposed three improvements: modified graph distances, distance-dependent weight matrices, and weighted sum conversion.
result The proposed method slightly outperforms the original weave module in compound activity prediction.
The paper refines the discourse on interpretability of machine learning models.
problem The ambiguity and diversity of motivations for interpretability in machine learning models.
method Examining motivations, model properties, and techniques thought to confer interpretability.
result Identifies transparency and post-hoc explanations as competing notions of interpretability.
GRAPE uses graph kernels to predict molecular energies efficiently.
problem Efficiently predicting molecular energies with physical constraints.
method GRAPE approach based on graph theory incorporating symmetries.
result GRAPE predicts atomization energies accurately on organic molecules.
Develops conformal Bayes for two-sided censored Gaussian regression under label shift.
problem Prediction under label shift with censored responses.
method Combines posterior predictive tilting with weighted conformal calibration.
result Restores marginal coverage with smaller prediction sets.
In many signal processing applications, the aim is to reconstruct a signal that has a simple representation with respect to a certain basis or frame. Fundamental elements of the basis known as "atoms" allow us to define "atomic norms" that can be used to formulate convex regularizations for the reconstruction problem. …
Geometric GNNs model 3D atomic systems with rotations and translations.
problem Modeling 3D atomic systems with geometric graphs and machine learning.
method Invariant, equivariant, and unconstrained GNN architectures.
result Geometric GNNs leverage physical symmetries and chemical properties.
Proposes an algorithm for infinite-dimensional sparse learning in system identification.
problem System identification without known model structures.
method Atomic norm regularization and greedy algorithm for solving an infinite-dimensional group lasso problem.
result The proposed algorithm outperforms benchmark methods in impulse response fitting and pole location estimation.
Optimizes ligand binding poses using CNNs and atomic grids.
problem Improving the accuracy of docking predictions for drug discovery.
method Differentiable atomic grid representation, CNN for scoring and optimization.
result Iteratively-trained CNNs outperform single CNNs in optimizing poses.
Generative models encode and decode 3D crystal structures from a large dataset.
problem Challenges in encoding and decoding 3D crystal structures from large datasets.
method Training two neural networks on a dataset of over 120,000 crystal structures to encode and decode 3D atom positions.
result Ability to generate compressed, continuous latent space representations and decode molecules accurately.
New method learns histograms using optimal transport barycenters.
problem Nonlinear dictionary learning for histograms.
method Optimal transport theory, displacement interpolations, entropic regularization, gradient descent.
result Efficient and tractable method for nonlinear dictionary learning.
New method for real-time reconstruction of sparse STEM images from non-rectangular scans.
problem Sparse sampling and non-rectangular scanning in STEM.
method General method for real-time reconstruction of sparsely sampled images from high-speed, non-invasive and diverse scanning pathways.
result Demonstrated on synthetic and experimental STEM data, achieving real-time reconstruction.
MuML models predict molecular dipole moments using atomic partial charges and dipoles.
problem Predicting molecular dipole moments accurately and efficiently.
method Combining atomic partial charges and atomic dipoles within a physically inspired ML model.
result MuML models achieve excellent transferability and accuracy, approaching DFT results at a fraction of the computational cost.
Optimizes basis for density-based atomic representations to enhance compactness and accuracy.
problem Improving the efficiency and accuracy of machine learning models for atomic properties.
method An unsupervised approach to determine the optimal basis set for atom density representations using splines.
result Optimal basis sets that encode structural information more compactly and accurately.
End-to-end model predicts protein interfaces from atomic coordinates.
problem Improving protein interface prediction using large datasets.
method Developed SASNet, an end-to-end learning model using only atomic coordinates.
result SASNet outperforms state-of-the-art methods trained on gold-standard data.
Recent research in off-the-grid compressed sensing (CS) has demonstrated that, under certain conditions, one can successfully recover a spectrally sparse signal from a few time-domain samples even though the dictionary is continuous. In particular, atomic norm minimization was proposed in \cite{tang2012csotg} to recove…
A new graph model HMG and neural network HMGNN improve molecule property predictions.
problem Predicting quantum mechanical properties of molecules with limited consideration of many-body interactions.
method Introducing heterogeneous molecular graphs (HMG) and building HMGNN on neural message passing scheme.
result HMGNN achieves state-of-the-art performance in 9 out of 12 tasks on the QM9 dataset.