The study reveals how attention paths in Transformers influence learning outcomes.
problem Understanding the theoretical basis of Transformers' performance.
method Developed a statistical mechanics theory for a simplified attention network.
result The predictor statistics are influenced by the combination of attention paths.
Proposes HetSANN for learning heterogeneous graph structures without meta-paths.
problem Learning low-dimensional vector space of heterogeneous information networks.
method Implicitly represents heterogeneous information through entity space transformation and attention mechanism.
result Significant improvements over state-of-the-art solutions on public datasets.
SigMA uses signatures and attention to estimate parameters in fBm-driven SDEs.
problem Estimating parameters in SDEs driven by fBm is challenging due to non-Markovian and semimartingale issues.
method SigMA integrates path signatures with multi-head self-attention, using convolutional and MLP layers.
result SigMA outperforms other methods in accuracy, robustness, and model compactness.
A new network learns to prioritize messages for efficient multi-robot path planning.
problem Efficient path planning and coordination for large-scale multi-robot systems.
method Message-Aware Graph Attention Network (MAGAT) incorporating attention mechanisms.
result MAGAT achieves performance close to a coupled centralized expert algorithm.
New method for learning on heterogeneous graphs without meta-paths.
problem Learning on heterogeneous graphs is sensitive to meta-paths choice, leading to poor performance.
method Decompose heterogeneous graph into homogeneous relation-type graphs, combine higher-order representations, use attention mechanisms.
result Our model outperforms state-of-the-art baselines in vertex classification tasks on heterogeneous graph datasets.
Transformers can solve complex filtering problems for non-Gaussian signals.
problem Non-linear and non-Markovian filtering problems for conditionally Gaussian signals.
method Continuous-time transformer models called filterformers.
result Filterformers can approximate the conditional law of non-Markovian and conditionally Gaussian signal processes.
Transformer model improves asset allocation by unifying forecasting and optimization.
problem Separation of forecasting and optimization leads to suboptimal portfolios.
method Signature Informed Transformer using path signatures and specialized attention.
result Direct minimization of Conditional Value at Risk improves performance.
A concise discussion of the axiomatic approach to the concept of parallel transport is presented. Attention is drawn to a bijective map between the sets of connections and (axiomatically defined) parallel transports. The transports along paths are pointed as a generalization of the (axiomatically defined) parallel tran…
Recently, several studies have explored methods for using KG embedding to answer logical queries. These approaches either treat embedding learning and query answering as two separated learning tasks, or fail to deal with the variability of contributions from different query paths. We proposed to leverage a graph attent…
This article reports on an investigation of the use of convolutional neural networks to predict the visual attention of chess players. The visual attention model described in this article has been created to generate saliency maps that capture hierarchical and spatial features of chessboard, in order to predict the pro…
In this paper we use a time-evolving graph which consists of a sequence of graph snapshots over time to model many real-world networks. We study the path classification problem in a time-evolving graph, which has many applications in real-world scenarios, for example, predicting path failure in a telecommunication netw…
Dupire's functional Itô calculus provides an alternative approach to the classical Malliavin calculus for the computation of sensitivities, also called Greeks, of path-dependent derivatives prices. In this paper, we introduce a measure of path-dependence of functionals within the functional Itô calculus framework. Name…
Paper proposes HIDAM model to improve MSE default risk assessment using heterogeneous information networks.
problem Default risk assessment for MSEs due to lack of credit information and diverse financial activities.
method HIDAM model incorporating heterogeneous information networks with multi-typed nodes and links, extracting interactive information through meta-paths, and using a hierarchical attention mechanism.
result HIDAM model outperforms state-of-the-art competitors on real-world banking data.
Hierarchical pretraining with slow-fast ODEs
problem Causal self-attention vs. slow-fast ODEs
method Instantiating fast-slow ODE formalism as a concrete neural network
result Equilibrium manifold x=φ(y) is exactly the master-equation (ME) stationary distribution Much of the recent work on learning molecular representations has been based on Graph Convolution Networks (GCN). These models rely on local aggregation operations and can therefore miss higher-order graph properties. To remedy this, we propose Path-Augmented Graph Transformer Networks (PAGTN) that are explicitly built…
We propose the Gaussian attention model for content-based neural memory access. With the proposed attention model, a neural network has the additional degree of freedom to control the focus of its attention from a laser sharp attention to a broad attention. It is applicable whenever we can assume that the distance in t…
Transformer adapts to graphs with adaptive attention and auto-regressive decoding.
problem Transformers struggle with graph data due to non-sequential nature.
method Proposes GRAT, a Transformer variant with adaptive attention and auto-regressive decoding.
result GRAT achieves state-of-the-art performance on molecule property predictions and generation tasks.
ABIForest improves anomaly detection using attention weights.
problem Anomaly detection in datasets.
method Attention mechanism integrated into Isolation Forest.
result ABIForest outperforms standard Isolation Forest on synthetic and real datasets.
Neural A* uses machine learning to improve path planning efficiency.
problem Challenges in applying machine learning to search-based path planning.
method Reformulated A* search as a differentiable network coupled with a convolutional encoder.
result Neural A* outperforms state-of-the-art planners in optimality and efficiency.
Improved speech enhancement with MNTFA using time-frequency attention.
problem Speech enhancement with limited model size and memory.
method Designing MNTFA with self-attention modules for long sequences and joint training.
result MNTFA achieves better performance with fewer parameters than DPCRN.
Generative models use kernel smoothing for conditioning on small example sets.
problem Improving generative models' performance with limited conditioning examples.
method Showed that cross-attention conditioning is equivalent to kernel smoothing, specifically a Nadaraya--Watson kernel smoother.
result The approach predicts and confirms three failure regimes for kernel-based conditioning.
GAT-RWOS uses graph attention to improve imbalanced data classification.
problem Imbalanced data leads to biased models favoring majority classes.
method Combines GATs and random walks to generate synthetic minority samples.
result Improves classification performance on imbalanced datasets.
A major drawback of backpropagation through time (BPTT) is the difficulty of learning long-term dependencies, coming from having to propagate credit information backwards through every single step of the forward computation. This makes BPTT both computationally impractical and biologically implausible. For this reason,…
As adversarial attacks pose a serious threat to the security of AI system in practice, such attacks have been extensively studied in the context of computer vision applications. However, few attentions have been paid to the adversarial research on automatic path finding. In this paper, we show dominant adversarial exam…
Proposes PGPS for efficient Bayesian inference.
problem Efficient sampling from complex posterior distributions.
method Path-guided particle-based sampling with Log-weighted Shrinkage.
result PGPS generates samples closer to target distribution.
Paper tackles offline SSP with value iteration for policy evaluation and learning.
problem Goal-oriented RL with offline data and cost minimization.
method Simple value iteration algorithms for OPE and offline policy learning.
result Strong instance-dependent bounds implying near-minimax optimal worst-case bounds.
Heterogeneous information network (HIN) embedding has gained increasing interests recently. However, the current way of random-walk based HIN embedding methods have paid few attention to the higher-order Markov chain nature of meta-path guided random walks, especially to the stationarity issue. In this paper, we system…
A new system recommends specific knowledge concepts in MOOCs based on student interests.
problem MOOCs recommend courses but ignore specific knowledge concepts interests.
method End-to-end graph neural network (ACKRec) that combines content and context information.
result ACKRec effectively recommends knowledge concepts to MOOC students.
Without relevant human priors, neural networks may learn uninterpretable features. We propose Dynamics of Attention for Focus Transition (DAFT) as a human prior for machine reasoning. DAFT is a novel method that regularizes attention-based reasoning by modelling it as a continuous dynamical system using neural ordinary…
MAGNA improves graph neural networks by incorporating multi-hop context information.
problem Limited context in current graph neural networks.
method Diffuses attention scores across the network, accounting for all paths between nodes.
result State-of-the-art performance on node classification and knowledge graph completion benchmarks.
Generalized are the investigated in other works of the author transports along paths in fibre bundles to transports along arbitrary maps in them. Their structure and some properties are studied. Special attention is paid to the linear case and the case when the map's domain is a Cartesian product of two sets. Also cons…
ie-HGCN addresses HIN challenges by efficiently learning node representations.
problem Lack of flexibility in exploring meta-paths and high computational complexity in HIN GCN methods.
method Hierarchical aggregation architecture that automatically extracts useful meta-paths and reduces computational cost.
result ie-HGCN outperforms state-of-the-art methods on real network datasets.
Paper proposes a method to evaluate SME credit risk using meta paths.
problem Evaluate credit risk of small and medium-sized enterprises with limited data.
method Exploits the representative power of information networks and meta paths to infer SME financial status.
result Meta path feature effectively identifies SMEs with credit risks.
A new router uses attention-based reinforcement learning to solve detailed routing problems efficiently.
problem Solving detailed routing in integrated circuits while adhering to complex design rules.
method Attention-based reinforcement learning applied to track-assignment detailed routing.
result The attention router achieves over 100x acceleration compared to a genetic router without sacrificing solution quality.
A brain-inspired spiking Transformer reduces energy consumption and enhances interpretability.
problem Energy inefficiency and lack of interpretability in Transformer models.
method Spiking STDP Transformer using spike-timing-dependent plasticity (STDP) for self-attention.
result Achieves 94.35% and 78.08% accuracy on CIFAR-10 and CIFAR-100 datasets respectively, with 88.47% energy reduction.
The existing literature provides evidence that limit order book data can be used to predict short-term price movements in stock markets. This paper proposes a new neural network architecture for predicting return jump arrivals in equity markets with high-frequency limit order book data. This new architecture, based on …
A new method for name disambiguation in academic networks using multi-view attention and recurrent neural networks.
problem Disambiguating authors with the same name in large-scale academic networks.
method Multi-view Attention-based Pairwise Recurrent Neural Network (MA-PairRNN) that divides papers into blocks based on author attributes and merges blocks of the same author.
result MA-PairRNN significantly improves name disambiguation performance on real-world datasets.
New model forecasts stock market volatility better than existing methods.
problem Forecasting volatility in stock markets.
method Combines HAR model with path-dependent volatility models.
result HAR-PD model family outperforms basic HAR model family in volatility forecasting.
ST-SAN predicts flow with spatial-temporal dependencies using self-attention.
problem Challenges in predicting flow due to spatial-temporal dependencies.
method Spatial-Temporal Self-Attention Network (ST-SAN) that addresses temporal and spatial dependencies.
result Significant improvement in flow prediction accuracy (9% in inflow, 4% in outflow) compared to state-of-the-art methods.
Unified theory explains two failure modes of deep transformers and provides initialisation guidelines.
problem Two failure modes (rank collapse and entropy collapse) of self-attention layers in deep transformers.
method Analytical theory of signal propagation through deep transformers, using the Random Energy Model analogy.
result Simple algorithm to compute trainability diagrams for correct initialisation hyper-parameters.
Transformers learn multi-step reasoning through gradient descent.
problem Understanding how transformers solve symbolic multi-step reasoning tasks.
method Theoretical analysis of gradient descent dynamics and multi-phase training.
result Trained one-layer transformers can solve both backward and forward reasoning tasks with generalization guarantees.
New imputation strategies improve signature models for irregular time series.
problem Applying signature models to irregular time series requires continuous path construction.
method Characterized imputation as a problem, evaluated various strategies, proposed GP-PoM.
result Gaussian process adapters improve predictive performance and robustness.
Rough Transformers improve time series modeling with lower costs and better performance.
problem Inefficient modeling of irregularly sampled time series data.
method Signature patching for continuous-time representations, reducing computational costs.
result Rough Transformers outperform vanilla Transformers and Neural ODE models.
A new method extracts events and their arguments efficiently from text.
problem Efficiently extract event information from texts with long-range dependencies and associations.
method Graph Convolutional Networks with shortest dependency paths to capture syntactic relationships.
result Significant improvement over state-of-the-art methods.
This research develops a new model for cyber risk and insurance pricing.
problem Accurate calculation of aggregate losses in cyber insurance pricing.
method A path-based k-generation risk contagion model in a tree-shaped network structure.
result Explicit expressions for mean and variance of local loss on a single path.
The celebrated Sequence to Sequence learning (Seq2Seq) technique and its numerous variants achieve excellent performance on many tasks. However, many machine learning tasks have inputs naturally represented as graphs; existing Seq2Seq models face a significant challenge in achieving accurate conversion from graph form …
Unified theory of optimal transport for random measures.
problem Statistical uncertainty in optimal transport.
method Constructing L2 over Wasserstein space for random probability measures. result Unified treatment of random optimal transport and principled inference.
Machine learning models have become more and more complex in order to better approximate complex functions. Although fruitful in many domains, the added complexity has come at the cost of model interpretability. The once popular k-nearest neighbors (kNN) approach, which finds and uses the most similar data for reasonin…