DLPaper2Code: Auto-Generation of Code From Deep Learning Research Papers
Authors: Akshay Sethi, Anush Sankaran, Naveen Panwar, Shreya Khare, Senthil Mani
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To evaluate our approach, we create a simulated dataset with over 216, 000 valid deep learning design flow diagrams using a manually defined grammar. Experiments on the simulated dataset show that the proposed framework provide more than 93% accuracy in flow diagram content extraction. |
| Researcher Affiliation | Collaboration | Akshay Sethi IIIT Delhi Anush Sankaran, Naveen Panwar, Shreya Khare, Senthil Mani IBM Research AI |
| Pseudocode | No | The paper describes its process through figures and text but does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | we created an ar Xiv-like website9, as shown in Figure 7, where the papers, the corresponding design, and the generated source code is available. The system and DLPaper2Code feature is found at: https://darviz.mybluemix.net/ |
| Open Datasets | Yes | A model simulation starts with an Input layer, where there are four possible options MNIST, CIFAR, Image Net, IMDBText. |
| Dataset Splits | Yes | The whole dataset is split as 60% for training, 20% for validation, and 20% for testing, making it a total of 195, 296 images for training and validation, and 48, 824 images for testing. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions software like Keras, Caffe, Tensorflow, Theano, Torch, MXNet, DL4J, CNTK, Lasagne, PyTorch, scikit-learn, VGG19, and Tesseract, but it does not provide specific version numbers for these software dependencies, which are necessary for replication. |
| Experiment Setup | Yes | All the figures are resized to 224 224 and 4, 096 features (fc2) are extracted from a fully connected layer of a popular deep learning model VGG19 (Simonyan and Zisserman 2014) pre-trained on Image Net dataset. We have two classification levels: (i) Coarse classifier: a binary neural network (NNet) classifier trained on fc2 features of VGG19 model to classify if the figure contains a DL model or not, and (ii) Fine-grained classifier: a five class neural network classifier trained on fc2 features to identify the type of DL design, only for those figures classified positive by the coarse classifier. The binary NNet classifier used two hidden layers of size [1024, 256]. |