1 code implementation • 1 Feb 2024 • A. Emin Orhan, Wentao Wang, Alex N. Wang, Mengye Ren, Brenden M. Lake
These results suggest that important temporal aspects of a child's internal model of the world may be learnable from their visual experience using highly generic learning algorithms and without strong inductive biases.
1 code implementation • 7 Aug 2023 • A. Emin Orhan
We find that it is feasible to reach human-level object recognition capacity at sub-human scales of model size, data size, and image size, if these factors are scaled up simultaneously.
1 code implementation • 24 May 2023 • A. Emin Orhan, Brenden M. Lake
Young children develop sophisticated internal models of the world based on their visual experience.
1 code implementation • 30 Mar 2023 • A. Emin Orhan
In recognition experiments, we ask if the model can distinguish the seen example from a novel example; in recall experiments, we ask if the model can correctly recall the seen example when cued by a part of it; and in retention experiments, we periodically probe the model's memory for the original examples as the model is trained continuously with new examples.
1 code implementation • 27 Apr 2022 • A. Emin Orhan
Humans have a remarkably large capacity to store detailed visual information in long-term memory even after a single exposure, as demonstrated by classic experiments in psychology.
1 code implementation • 30 Sep 2021 • A. Emin Orhan
Finally, we show that larger models are harder to train from scratch and their generalization accuracy is lower when trained up to convergence on the relatively small SCAN and COGS datasets, but the benefits of large-scale pretraining become much clearer with larger models.
1 code implementation • 23 Sep 2021 • A. Emin Orhan
The exact values of these estimates are sensitive to some underlying assumptions, however even in the most optimistic scenarios they remain orders of magnitude larger than a human lifetime.
1 code implementation • NeurIPS 2020 • A. Emin Orhan, Vaibhav V. Gupta, Brenden M. Lake
Within months of birth, children develop meaningful expectations about the world around them.
1 code implementation • 17 Jul 2019 • A. Emin Orhan
We show that these models display an unprecedented degree of robustness against common image corruptions and perturbations, as measured by the ImageNet-C and ImageNet-P benchmarks.
1 code implementation • 20 Jun 2019 • A. Emin Orhan, Brenden M. Lake
As reported in previous work, we show that an explicit episodic memory improves the robustness of image recognition models against small-norm adversarial perturbations under some threat models.
1 code implementation • ICLR 2020 • A. Emin Orhan, Xaq Pitkow
In the presence of a non-linearity, orthogonal transformations no longer preserve norms, suggesting that alternative transformations might be better suited to non-linear networks.
1 code implementation • NeurIPS 2018 • A. Emin Orhan
We propose to extract this extra class-relevant information using a simple key-value cache memory to improve the classification performance of the model at test time.
no code implementations • ICLR 2018 • A. Emin Orhan, Xaq Pitkow
Here, we present a novel explanation for the benefits of skip connections in training very deep networks.
1 code implementation • 12 Jan 2016 • A. Emin Orhan, Wei Ji Ma
We show that generic neural networks trained with a simple error-based learning rule perform near-optimal probabilistic inference in nine common psychophysical tasks.