HOUDINI: Lifelong Learning as Program Synthesis
Authors: Lazar Valkov, Dipak Chaudhari, Akash Srivastava, Charles Sutton, Swarat Chaudhuri
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate HOUDINI on three benchmarks that combine perception with the algorithmic tasks of counting, summing, and shortest-path computation. Our experiments show that HOUDINI transfers high-level concepts more effectively than traditional transfer learning and progressive neural networks, and that the typed representation of networks significantly accelerates the search. |
| Researcher Affiliation | Collaboration | Lazar Valkov University of Edinburgh L.Valkov@sms.ed.ac.uk; Dipak Chaudhari Rice University dipakc@rice.edu; Akash Srivastava University of Edinburgh Akash.Srivastava@ed.ac.uk; Charles Sutton University of Edinburgh, The Alan Turing Institute, and Google Brain charlessutton@google.com; Swarat Chaudhuri Rice University swarat@rice.edu |
| Pseudocode | No | The paper describes the learning algorithm and strategies in text but does not include structured pseudocode or an algorithm block. |
| Open Source Code | Yes | The implementation for HOUDINI is available online [1]. [1] Houdini code repository. https://github.com/capergroup/houdini. |
| Open Datasets | Yes | These tasks include object recognition tasks over three data sets: MNIST [21], NORB [22], and the GTSRB data set of images of traffic signs [37]. |
| Dataset Splits | No | The paper states that each task has its own training and validation set, and that evaluation is done on a validation set, but it does not specify the exact proportions or sample counts for these splits (e.g., '80/10/10 split'). |
| Hardware Specification | Yes | Experiments were performed using a single-threaded implementation on a Linux system, with 8-core Intel E5-2620 v4 2.10GHz CPUs and TITAN X (Pascal) GPUs. |
| Software Dependencies | No | The paper mentions general differentiable programming languages and frameworks in related work but does not explicitly list the specific software dependencies and their version numbers (e.g., Python, PyTorch/TensorFlow versions) used for their implementation. |
| Experiment Setup | No | The paper describes the general experimental design, including types of neural modules, synthesis strategies, and baselines used, but it does not specify concrete hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed system-level training settings. |