Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Learning Multi-Way Relations via Tensor Decomposition With Neural Networks
Authors: Koji Maruhashi, Masaru Todoriki, Takuya Ohwa, Keisuke Goto, Yu Hasegawa, Hiroya Inakoshi, Hirokazu Anai
AAAI 2018 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments on three different domains show that our method is highly accurate, especially on higher order data. It also enables us to interpret the classification results along with the matrices calculated with the novel tensor decomposition. |
| Researcher Affiliation | Industry | Koji Maruhashi, Masaru Todoriki, Takuya Ohwa, Keisuke Goto, Yu Hasegawa, Hiroya Inakoshi, Hirokazu Anai Fujitsu Laboratories Ltd. 4-1-1, Kamikodanaka, Nakahara-ku, Kawasaki, Kanagawa, Japan EMAIL |
| Pseudocode | Yes | Algorithm 1: Deep Tensor |
| Open Source Code | No | The paper mentions that their method is implemented in Python using Chainer, but does not provide a link or statement about their own source code being publicly available. |
| Open Datasets | Yes | We use DARPA Intrusion Detection Data Sets1 and use a 1998 dataset... 1http://www.ll.mit.edu/ideval/data/; transactions from the Show Me the Money website2... 2http://smtm.labs.theodi.org/; ENZYMES, NCI1, and NCI109, downloaded from the ETH z urich website 3... 3https://www.bsse.ethz.ch/mlcb/research/machinelearning/graph-kernels/weisfeiler-lehman-graph-kernels.html; CHEM is provided by the Pub Chem Bio Assay 4... 4https://www.ncbi.nlm.nih.gov/pcassay |
| Dataset Splits | Yes | We choose the best parameters by using the validation datasets... On BIO, we evaluate the methods based on 10-fold cross validation repeated five times and take their average accuracy. |
| Hardware Specification | Yes | All the experiments were conducted on an Intel(R) Xeon(R) CPU E5-2630 v3 2.40GHz with 132GB of memory and a Tesla P100 GPU, running Linux. |
| Software Dependencies | Yes | Deep Tensor and Tucker NN are implemented in Python 2.7, using Chainer 1.24 (http://chainer.org/). |
| Experiment Setup | Yes | The best size of the core tensor is mainly 10 10 10 10 for IDS, 10 10 10 for LEND, and 50 50 for BIO, chosen out of {10, 25, 50} for each mode. The best number of hidden layers is mainly 4 for IDS, 3 for LEND, 4 for ENZYMES, and 4 for NCI1 and NCI109, chosen out of various numbers of hidden layers; up to 7, with 1, 000 neurons in each layer. The number of epochs for training is 200 and the mini-batch size is 100. We use ReLU (Glorot, Bordes, and Bengio 2011) as the activation function and use it with or without dropout (Srivastava et al. 2014) and batch normalization (Ioffe and Szegedy 2015) techniques. |