Novel Dynamic Batch-Sensitive Adam Optimiser for Vehicular Accident Injury Severity Prediction

2026-05-14 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

Researchers Daniel Asare Kyei, Alimatu Saadia-Yussiff, Maame G. Asante-Mensah, Abdul Lateef-Yussiff, and Charles Roland Haruna introduce Dynamic Batch-Sensitive Adam (DBS-Adam), a novel optimizer designed to enhance deep learning model efficiency and convergence, particularly with imbalanced and sequential datasets. DBS-Adam dynamically adjusts the learning rate based on a batch difficulty score, which is calculated from exponential moving averages of gradient norms and batch loss. This mechanism stabilizes training and speeds up convergence by applying larger updates to challenging batches and smaller updates to easier ones. The optimizer was evaluated by integrating it into Bi-Directional LSTM networks for vehicular accident injury severity prediction, utilizing SMOTE-ENN resampling and Focal Loss to manage class imbalance. DBS-Adam achieved a 95.22% test accuracy, 96.11% precision, 95.28% recall, 95.39% F1-score, and a test loss of 0.0086, outperforming AMSGrad, AdamW, and AdaBound with statistically significant precision improvements (p=0.020).

Key takeaway

For AI Engineers developing models for imbalanced or sequential data, adopting DBS-Adam can significantly improve training stability and convergence speed. Your models, especially those using Bi-Directional LSTMs for tasks like real-time accident severity prediction, could achieve higher precision and overall performance. Consider integrating DBS-Adam alongside techniques like SMOTE-ENN and Focal Loss to maximize effectiveness on challenging datasets.

Key insights

DBS-Adam dynamically scales learning rates based on batch difficulty, improving deep learning performance on imbalanced, sequential data.

Principles

• Optimizers must adapt to data imbalance.
• Dynamic learning rates improve convergence.
• Batch difficulty scores guide optimization.

Method

DBS-Adam computes a batch difficulty score from exponential moving averages of gradient norms and batch loss, then dynamically scales the learning rate to increase updates for difficult batches and reduce them for easier ones.

In practice

• Integrate DBS-Adam with Bi-Directional LSTMs.
• Apply SMOTE-ENN for class imbalance.
• Use Focal Loss for imbalanced classification.

Topics

• Dynamic Batch-Sensitive Adam
• Vehicular Accident Prediction
• Bi-Directional LSTM
• Imbalanced Data Learning
• Focal Loss

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• Novel Dynamic Batch-Sensitive Adam Optimiser for Vehicular Accident Injury Severity Prediction — DBS-Adam dynamically scales learning rates based on batch difficulty, improving stability and convergence for imbalanced sequential data.
• Asymmetric-Loss-Guided Hybrid CNN-BiLSTM-Attention Model for Industrial RUL Prediction with Interpretable Failure Heatmaps — A hybrid CNN-BiLSTM-Attention model with asymmetric loss improves turbofan RUL prediction and interpretability.
• Scaling Learning-based AEB with Massive Unlabeled Data — Stabilized meta-feedback semi-supervised learning effectively scales AEB using massive unlabeled data, significantly improving safety metrics in real-world deployment.
• Singularity-aware Optimization via Randomized Geometric Probing: Towards Stable Non-smooth Optimization — S-Adam stabilizes non-smooth deep learning optimization by dynamically adjusting step sizes based on local geometric instability, preventing gradient chattering.
• LLMs on Tabular Data with Limited Semantics: Evidence from Industrial Car Retrofit Prediction — LLMs serve best as complementary components to strong tabular machine learning baselines, not replacements, for industrial data with limited semantics.
• Optimizers in Deep Learning: From Gradient Descent to Adam — Optimizers guide neural network training by efficiently updating parameters to minimize loss, evolving from basic GD to adaptive Adam.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.