lstm

The paper proposes MambaLSTM, a framework combining Mamba state-space models and LSTM for spatio-temporal traffic accident risk prediction, addressing noise in feature fusion and global spatial correlation.

This paper proposes a dual-domain fused LSTM (DDF-LSTM) model for time-dependent reliability analysis that integrates time-independent random variables and stochastic processes to efficiently estimate failure probabilities via Monte Carlo simulation.

This paper proposes PTTSD, a probabilistic framework for depression severity detection from clinical interview transcripts that models uncertainty and provides temporal interpretability, achieving competitive performance on benchmark datasets.

This paper presents a comparative study of LSTM and traditional models for sentiment analysis of public opinion.

This paper proposes INTENT, an LSTM framework for predicting vehicle intentions (straight, left turn, right turn) at intersections 2 seconds ahead, achieving 99.71% accuracy on the InD dataset.

This paper evaluates the effect of frame sampling rate on sequence-based classification of autism-related self-stimulatory hand behaviors using LSTM and GRU models, achieving up to 98.75% accuracy at a 15-frame interval, and analyzes data augmentation strategies for small behavioral video datasets.

Proposes an uncertainty-aware reinforcement learning framework for algorithmic trading that integrates distributional, epistemic, and aleatoric uncertainty using SHAP-weighted reconstruction, MC Dropout, and LSTM consensus. Outperforms traditional models on five major US stock indices.

This paper proposes a data-driven surrogate modeling framework using a hybrid Graph Neural Network-Long Short-Term Memory architecture to predict the static response of additively manufactured short-fiber thermoplastics, achieving high accuracy (R²≈0.98) and two orders of magnitude speedup over finite element simulations.

This study investigates machine learning models to predict exam outcomes using physiological data such as electrodermal activity, heart rate, and skin temperature, finding that both deep learning approaches and simpler models like random forests can be effective.

This white paper proposes using LSTM neural networks to detect structural breaks in property insurance loss reserving caused by climate-driven catastrophes, aiming to improve accuracy by 15–20% over traditional methods like Chain Ladder.

GlucoFM-Bench evaluates time-series foundation models for blood glucose forecasting across 15 datasets, showing strong zero-shot/few-shot transfer by Chronos-2 and TimesFM but superior performance of a lightweight LSTM when full training data is available.

Researchers propose a lightweight autoregressive framework for graph generation that uses structure-guided topological ordering to achieve near log-linear complexity, addressing scalability and novelty limitations of existing diffusion and autoregressive methods. The approach supports both LSTM and Mamba-style backbones and shows improved novelty while maintaining validity and uniqueness on molecular and non-molecular benchmarks.

Researchers propose a Physics-Informed Machine Learning (PIML) framework that integrates hydrological constraints into an LSTM loss function to improve short-term flood forecasting, particularly in data-scarce regimes. A 'Trend Alignment' constraint enforcing consistency between precipitation and discharge trends improves Nash-Sutcliffe Efficiency and eliminates unphysical predictions during extreme events.

This paper evaluates encoder-only Transformer and LSTM models for streamflow prediction in ungauged basins using NOAA's National Water Model simulations. Results show LSTM outperforms Transformer, and incorporating downstream information significantly improves prediction skill across both architectures.

This paper uses a BERT-based large language model for sentiment analysis of Decentraland's Discord community to enhance MANA token price prediction, demonstrating that a multi-modal LSTM incorporating sentiment, trading volume, and market capitalization outperforms a price-only baseline.

OpenAI proposes Teacher–Student Curriculum Learning (TSCL), a framework where a Teacher algorithm automatically selects subtasks for a Student to learn complex tasks, optimizing based on learning curve slope and preventing forgetting. The approach matches or surpasses hand-crafted curricula on decimal addition and Minecraft navigation tasks, enabling solutions previously impossible with direct training.

OpenAI demonstrates an unsupervised system that learns sentiment representation by training a multiplicative LSTM to predict the next character in Amazon reviews, achieving state-of-the-art sentiment analysis on Stanford Sentiment Treebank (91.8% accuracy) while requiring 30-100x fewer labeled examples than supervised approaches. The model discovers a distinct 'sentiment neuron' that captures sentiment information and can be directly manipulated to control text generation sentiment.