When to Intervene: Learning Optimal Intervention Policies for Critical Events
Authors: Niranjan Damera Venkata, Chiranjib Bhattacharyya
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we demonstrate RNN-based OTI policies with experiments and show that they outperform popular intervention methods. and Experiments are performed on two real-world datasets |
| Researcher Affiliation | Collaboration | Niranjan Damera Venkata Digital and Transformation Organization HP Inc., Chennai, India niranjan.damera.venkata@hp.com Chiranjib Bhattacharyya Dept. of CSA and RBCCPS Indian Institute of Science, Bangalore, India chiru@iisc.ac.in |
| Pseudocode | No | The paper includes an architecture diagram (Figure 1) but does not contain any pseudocode or algorithm blocks. |
| Open Source Code | No | The code is proprietary. |
| Open Datasets | Yes | Turbofan Engine Failure Data [26, 32]: This dataset from NASA provides train and test data... Azure Predictive Maintenance Guide Data[25]: This is a dataset from a guide provided by Microsoft |
| Dataset Splits | Yes | For both datasets, we train on 70% of randomly selected co-variate time-series sequences and hold out 30% of the sequences for testing. From the training set a further 30% of the sequences are set aside as validation data to tune model parameters and policy thresholds. This process is repeated to produce 10 random train-validation-test splits. |
| Hardware Specification | Yes | All experiments were run on a Tensorbook Laptop with 32GB RAM and having a single NVIDIA Ge Force GTX 1070 with Max-Q GPU. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software dependencies used in the experiments (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | All RNNs share the same encoder architecture which is an LSTM with 128 step look-back and hidden state dimension of 16 units. WBI and TTE thresholds are tuned (individually, for 125 different settings of Cα {8, 10, , 256}) on each validation set using an empirical intervention policy risk: |