Hybrid LSTM and PPO Networks for Dynamic Portfolio Optimization

A recent study compared the performance of Kolmogorov-Arnold Networks (KAN) and Long Short-Term Memory (LSTM) networks in forecasting non-deterministic stock price data. The findings revealed that LSTM outperformed KAN across all tested prediction horizons, demonstrating its effectiveness in sequential data modeling while KAN showed higher error rates despite its theoretical interpretability.

A new reinforcement learning platform named ContagionRL has been introduced, designed for reward engineering in spatial epidemic simulations. This platform allows researchers to evaluate how different reward function designs influence survival strategies in various epidemic scenarios, integrating a spatial SIRS+D epidemiological model with adjustable environmental parameters.

A recent study has introduced a bidirectional Long Short-Term Memory (LSTM) neural network designed to classify light curves of transient astronomical objects using data from the Photometric LSST Astronomical Time-series Classification Challenge (PLAsTiCC). The study reorganized the original fourteen object classes into five categories to mitigate class imbalance, achieving high performance in certain classifications while facing challenges with others.

A recent study explores the enhancement of Reinforcement Learning (RL) in 3D visuospatial tasks through a human-informed curriculum design, aiming to improve the technology's effectiveness in complex problem domains. The research highlights the challenges faced by state-of-the-art RL methods, such as PPO and imitation learning, in mastering these tasks.

A new framework for Electric Vehicle (EV) charging forecasting has been proposed, addressing cybersecurity threats that compromise operational efficiency and grid stability. This framework employs a federated learning approach that integrates LSTM autoencoder-based anomaly detection, interpolation for data mitigation, and collaborative learning without centralized data aggregation.

A new algorithm, Deep Gaussian Process Proximal Policy Optimization (GPPO), has been introduced to enhance uncertainty estimation in Reinforcement Learning (RL), particularly in control tasks requiring a balance between safe exploration and efficient learning. GPPO utilizes Deep Gaussian Processes to approximate both policy and value functions, maintaining competitive performance with existing methods while offering calibrated uncertainty estimates.

A new dataset named MultiBanAbs has been developed to facilitate Bangla abstractive summarization, comprising over 54,000 articles and summaries from diverse sources including blogs and newspapers. This initiative addresses the limitations of existing summarization systems that primarily focus on news articles, which often do not reflect the varied nature of real-world Bangla texts.

Recent research has introduced Liquid Time-Constant Networks (LTCs) and Closed-form Continuous-time Networks (CfCs) to model retinal ganglion cell activity in tiger salamanders, demonstrating lower mean absolute error (MAE) and faster convergence compared to traditional convolutional models and LSTMs.