Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs
Authors: Jonas Kubilius, Martin Schrimpf, Kohitij Kar, Rishi Rajalingham, Ha Hong, Najib Majaj, Elias Issa, Pouya Bashivan, Jonathan Prescott-Roy, Kailyn Schmidt, Aran Nayebi, Daniel Bear, Daniel L. Yamins, James J. DiCarlo
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We developed CORnet-S, a shallow ANN with four anatomically mapped areas and recurrent connectivity, guided by Brain-Score, a new large-scale composite of neural and behavioral benchmarks for quantifying the functional fidelity of models of the primate ventral visual stream. Despite being significantly shallower than most models, CORnet-S is the top model on Brain-Score and outperforms similarly compact models on Image Net. |
| Researcher Affiliation | Collaboration | 1Mc Govern Institute for Brain Research, MIT, Cambridge, MA 02139 2Brain and Cognition, KU Leuven, Leuven, Belgium 3Department of Brain and Cognitive Sciences, MIT, Cambridge, MA 02139 4Center for Brains, Minds and Machines, MIT, Cambridge, MA 02139 5Bay Labs Inc., San Francisco, CA 94102 6Center for Neural Science, New York University, New York, NY 10003 7Department of Neuroscience, Zuckerman Mind Brain Behavior Institute, Columbia University, New York, NY 10027 8Neurosciences Ph D Program, Stanford University, Stanford, CA 94305 9Department of Psychology, Stanford University, Stanford, CA 94305 10Department of Computer Science, Stanford University, Stanford, CA 94305 |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Image Net-pretrained CORnet-S is available at github.com/dicarlolab/cornet. |
| Open Datasets | Yes | We used Py Torch 0.4.1 and trained the model using Image Net 2012 [34]... A total of 1318 grayscale images, containing images from Section 3.1 and MS COCO [26]... we tested how well these models generalize to CIFAR-100 dataset |
| Dataset Splits | Yes | Images were preprocessed (1) for training random crop to 224 224 pixels and random flipping left and right; (2) for validation central crop to 224 224 pixels... 40,000 images from CIFAR-100 train set were used for training and the remaining 10,000 for testing |
| Hardware Specification | Yes | trained on 2 GPUs (NVIDIA Titan X / Ge Force 1080Ti) |
| Software Dependencies | Yes | We used Py Torch 0.4.1 and trained the model... We used a scikit-learn implementation of a multinomial logistic regression using L-BFGS [31] |
| Experiment Setup | Yes | We used a batch size of 256 images and trained on 2 GPUs (NVIDIA Titan X / Ge Force 1080Ti) for 43 epochs. We use similar learning rate scheduling to Res Net with more variable learning rate updates (primarily in order to train faster): 0.1, divided by 10 every 20 epochs. For optimization, we use Stochastic Gradient Descent with momentum .9, a cross-entropy loss between image labels and model predictions (logits). |