Reconstructing perceived faces from brain activations with deep adversarial neural decoding

Authors: Yağmur Güçlütürk, Umut Güçlü, Katja Seeliger, Sander Bosch, Rob van Lier, Marcel A. J. van Gerven

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our approach with a functional magnetic resonance imaging experiment and show that it can generate state-of-the-art reconstructions of perceived faces from brain activations.
Researcher Affiliation Academia Ya˘g mur Güçlütürk*, Umut Güçlü*, Katja Seeliger, Sander Bosch, Rob van Lier, Marcel van Gerven, Radboud University, Donders Institute for Brain, Cognition and Behaviour Nijmegen, the Netherlands {y.gucluturk, u.guclu}@donders.ru.nl
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or link for open-source code for the described methodology. It mentions that its fMRI dataset is 'available from the first authors on reasonable request,' but this refers to data, not code.
Open Datasets Yes Celeb A dataset [48]. This dataset comprises 202599 in-the-wild portraits of 10177 people, which were drawn from online sources.
Dataset Splits Yes In total, 700 faces were presented twice for the training set, and 48 faces were repeated 13 times for the test set.
Hardware Specification Yes This work has been partially supported by a VIDI grant (639.072.513) from the Netherlands Organization for Scientific Research and a GPU grant (Ge Force Titan X) from the Nvidia Corporation.
Software Dependencies No The paper mentions several software components like 'Chainer and Cupy with CUDA and cu DNN', 'Caffe', 'scikit-learn', 'SPM', 'FSL', and 'Adam', but it does not provide specific version numbers for any of these components.
Experiment Setup Yes Model parameters were initialized as follows: biases were set to zero, the scaling parameters were drawn from N(1, 2 · 10−2I), the shifting parameters were set to zero and the weights were drawn from N(1, 10−2I) [37]. We set the hyperparameters of the loss functions as follows: λadv = 102, λdis = 102, λfea = 10−2 and λsti = 2 · 10−6 [38]. We set the hyperparameters of the optimizer as follows: α = 0.001, β1 = 0.9, β2 = 0.999 and ϵ = 108 [37].