2022年9月24日 · 为了准确和及时预测局部区域的降雨及温度，提出了一种基于 Attention 和 LSTM 组合模型 ( ALSTM) 的 关联多值 预测算法。 该算法利用天气时间序列中 的前期数据，对下一小时的降雨量和温度进行关联预测，以此实现对天气要素的多值预测。 该算法首先对输入数据进行归一化处理; 然后利用数据对ALSTM 模型进行训练; 最后将训练好的模型用于多值预测。 涉及：LSTM Attention机制 多变量. ③将两个或多个LSTM模型进行 堆叠组合，将前一个模型的输出作为下 …

跑完原生的ALSTM Benchmark发现不考虑交易成本的净值曲线居然近似相当于基准的沪深300指数（即几乎无超额收益），而考虑了交易成本的超额收益居然是负的，观察前面模型训练的每个epoch的输出发现train set的loss一直在下降，而valid的loss却缓慢上升。

1.预测模型：ALSTM-Attentive LSTM. 先看整体结构： 和标准的lstm相比，在上面的倒数第二层增加了“ temporal attention layer ”，也就是说，在预测前，综合使用了所有时刻隐藏层，并给它们不同的权重。这和之前看过的 DARNN 的时态注意力机制并没什么不同。 具体的做法：

In this paper, we propose ALSTM model, which is based on the Long Short Term Memory (LSTM) recurrent neural network and the attention mechanism for multi-scenario bandwidth prediction. Firstly, we conduct the bandwidth trajectories feature analysis, and then we adopt the Support Vector Machine (SVM) to classify scenarios based on the bandwidth ...

其实本文模型的原型是ALSTM (Attentive LSTM), 是机器学习选股常见的，也是微软qlib自带的一个baseline，只不过本文作者为了降低模型的过拟合并提高泛化能力，在每一次Iteration里面增添了对抗样本 e_ {adv} 以及对应的损失函数部分。 我对网络的完整架构和单次BP的过程做了手动复现。 由于未参考源代码，或存在部分不妥之处欢迎交流指正。 原论文：《Enhancing Stock Movement Prediction with Adversarial Training》

2020年7月25日 · In this paper, we produce a novel framework (ALSTM) based on the Attention mechanism and Long Short-Term Memory (LSTM), which associates structure learning with parameter learning of first-order logical rules in an end-to …

To run the proposed Adv-ALSTM, ALSTM, and LSTM on the ACL18 and KDD17 dataset, run the commands in the hyperparameter file.

This paper proposes a novel attention-based long short-term memory (ALSTM) method for improved design of GC models to diagnose the alarm root-cause and identify the alarm propagation pathways. A novel training strategy is proposed for ALSTM-GC to enhance noise robustness and prevent overfitting.

PyTorch implementation of the adaptive LSTM (https://arxiv.org/abs/1805.08574), an extension of the standard LSTM that increases model flexibility through adaptive parameterization. The aLSTM converges faster than the LSTM with superior generalizing performance.

The project is all about reproducing a working version of Adversarial Attention Based LSTM for TensorFlow 2. This new version is available in AdvALSTM.py with the object AdvLSTM. More details of my version below. I finally updated the author's code for it to run with TensorFlow 2.x and compare my results with his.