Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks
Authors: Pouya Bashivan, Irina Rish, Mohammed Yeasin, Noel Codella
ICLR 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical evaluation on the cognitive load classification task demonstrated significant improvements in classification accuracy over current state-of-the-art approaches in this field. reducing the classification error from 15.3% (state-of-art on this application) to 8.9%. |
| Researcher Affiliation | Collaboration | Pouya Bashivan Electrical and Computer Engineering Department University of Memphis Memphis, TN , USA {pbshivan}@memphis.edu Irina Rish IBM T.J. Watson Research Center Yorktown Heights, NY, USA {rish}@us.ibm.com Mohammed Yeasin Electrical and Computer Engineering Department University of Memphis Memphis, TN , USA {myeasin}@memphis.edu Noel Codella IBM T.J. Watson Research Center Yorktown Heights, NY, USA {nccodell}@us.ibm.com |
| Pseudocode | No | No pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | Yes | The code necessary for generating EEG images and building and training the networks discussed in this paper is available online2. 2https://github.com/pbashivan/EEGLearn |
| Open Datasets | No | The paper mentions using an EEG dataset acquired during an experiment and reports details in a previous publication (Bashivan et al., 2014), but it does not provide a direct link, DOI, repository name, or explicit statement of public availability for the dataset itself. |
| Dataset Splits | Yes | For evaluating the performance of each classifier we followed the leave-subject-out cross validation approach. In each of the 13 folds, all trials belonging to one of the subjects were used as the test set. A number of samples equal to the test set were then randomly extracted from rest of data for validation set and the remaining samples were used as training set. |
| Hardware Specification | No | No specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments were provided in the paper. |
| Software Dependencies | No | The paper mentions using 'Lasagne' and 'Adam algorithm' but does not specify version numbers for these software components. |
| Experiment Setup | Yes | We trained the recurrent-convolutional network with Adam algorithm (Kingma & Ba, 2015) with a learning factor of 10 3, and decay rate of first and second moments as 0.9 and 0.999 respectively. Batch size was set to 20. 50% dropout was used on the last two fully connected layers. The network parameters converge after about 600 iterations (5 epochs). |