Dynamic Inference with Neural Interpreters
Authors: Nasim Rahaman, Muhammad Waleed Gondal, Shruti Joshi, Peter Gehler, Yoshua Bengio, Francesco Locatello, Bernhard Schölkopf
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we empirically evaluate Neural Interpreters in a variety of problem settings. |
| Researcher Affiliation | Collaboration | 1Max-Planck-Institute for Intelligent Systems, Tübingen. 2Mila, Québec. 3Amazon Web Services. Work partially done during an internship at Amazon Web Services. |
| Pseudocode | No | The paper describes the architecture and its components using mathematical equations and text, but does not include any explicit pseudocode blocks or algorithms labeled as such. |
| Open Source Code | No | The paper does not contain an explicit statement about the release of source code or a link to a code repository. |
| Open Datasets | Yes | We consider three related datasets sharing the same label semantics, namely SVHN [41], MNISTM [20] and MNIST [35]... In addition, we also use unaugmented samples from the K-MNIST dataset [12] of Hiragana characters to probe fast-adaptation to new data. |
| Dataset Splits | Yes | For each function, we sample 163840 points from a uniform distribution over [0, 1]5, of which we use 80% for training, and the remaining for validation. |
| Hardware Specification | No | The paper does not specify any particular hardware components (e.g., GPU, CPU models, or cloud computing instances with detailed specifications) used for conducting the experiments. |
| Software Dependencies | No | The paper mentions software like 'Pytorch image models' [60] and 'GELUs' [26] (an activation function) and frameworks like PyTorch, but does not provide specific version numbers for any software dependencies or libraries used in the experiments. |
| Experiment Setup | Yes | We pre-train a Neural Interpreter to regress all 20 functions simultaneously... train their outputs to regress the respective function. Having pre-trained the model for 20 epochs, we finetune it for an additional 3 epochs... We train Neural Interpreters for 100 epochs on the combined digits dataset... we finetune the model on K-MNIST with varying numbers of samples for 10 epochs. |