Letter perception emerges from unsupervised deep learning and recycling of natural image features

@article{Testolin2017LetterPE,
  title={Letter perception emerges from unsupervised deep learning and recycling of natural image features},
  author={Alberto Testolin and Ivilin Stoianov and Marco Zorzi},
  journal={Nature Human Behaviour},
  year={2017},
  volume={1},
  pages={657-664}
}
The use of written symbols is a major achievement of human cultural evolution. However, how abstract letter representations might be learned from vision is still an unsolved problem1,2. Here, we present a large-scale computational model of letter recognition based on deep neural networks3,4, which develops a hierarchy of increasingly more complex internal representations in a completely unsupervised way by fitting a probabilistic, generative model to the visual input5,6. In line with the… Expand
Learning representation hierarchies by sharing visual features: a computational investigation of Persian character recognition with unsupervised deep learning
TLDR
A computational model of Persian character recognition based on deep belief networks is presented, where increasingly more complex visual features emerge in a completely unsupervised manner by fitting a hierarchical generative model to the sensory data. Expand
General object-based features account for letter perception better than specialized letter features
TLDR
Behavioral-computational evidence is provided that the perception of letters depends on general visual features rather than a specialized feature space, and several approaches to alter object-based features with letter specialization did not improve the match to human behavior. Expand
Unsupervised learning predicts human perception and misperception of gloss
TLDR
Linearly decoding specular reflectance from the model’s internal code predicts human gloss perception better than ground truth, supervised networks, or control models, and predicts, on an image-by-image basis, illusions of gloss perception caused by interactions between material, shape, and lighting. Expand
Understanding Character Recognition using Visual Explanations Derived from the Human Visual System and Deep Networks
TLDR
This work uses eye-tracking to assay the spatial distribution of information hotspots for humans via fixation maps and an activation mapping technique for obtaining analogous distributions for deep networks through visualization maps and finds an interesting correlate of congruence. Expand
An emergentist perspective on the origin of number sense
  • M. Zorzi, Alberto Testolin
  • Psychology, Medicine
  • Philosophical Transactions of the Royal Society B: Biological Sciences
  • 2018
TLDR
It is shown that deep neural networks endowed with basic visuospatial processing exhibit a remarkable performance in numerosity discrimination before any experience-dependent learning, whereas unsupervised sensory experience with visual sets leads to subsequent improvement of number acuity and reduces the influence of continuous visual cues. Expand
Deep learning systems as complex networks
TLDR
This article proposes to study deep belief networks using techniques commonly employed in the study of complex networks, in order to gain some insights into the structural and functional properties of the computational graph resulting from the learning process. Expand
Learning Numerosity Representations with Transformers: Number Generation Tasks and Out-of-Distribution Generalization
TLDR
It is shown that attention-based architectures operating at the pixel level can learn to produce well-formed images approximately containing a specific number of items, even when the target numerosity was not present in the training distribution. Expand
Do Humans and Deep Convolutional Neural Networks Use Visual Information Similarly for the Categorization of Natural Scenes?
TLDR
If the visual characteristics of the environment are not learned by CNNs, their visual categorization may depend only on a subset of the visual information on which humans rely, for example, on low spatial frequency information. Expand
Publisher Correction: Letter perception emerges from unsupervised deep learning and recycling of natural image features
TLDR
In the version of this Letter originally published, in the sentence beginning “Written symbols are culture specific...”, ‘Φ’ was used instead of ‘F’; it should have read ‘(for example, ℱ versus F)’. Expand
Representation of white- and black-box adversarial examples in deep neural networks and humans: A functional magnetic resonance imaging study
TLDR
Comparing the visual representations of white- and black-box adversarial examples in DNNs and humans by leveraging functional magnetic resonance imaging (fMRI) suggests that adversarialExamples may be differentially represented in the human visual system, but unable to affect the perceptual experience. Expand
...
1
2
3
...

References

SHOWING 1-10 OF 90 REFERENCES
Deep generative learning of location-invariant visual word recognition
TLDR
The results reveal that the efficient coding of written words—which was the model's learning objective—is largely based on letter-level information. Expand
Adaptation of the human visual system to the statistics of letters and line configurations
TLDR
A reorganization of the visual pathway resulting from reading acquisition under genetic and connectional constraints is suggested, suggesting an early visual tuning to large visual features such as letters. Expand
Emergence of simple-cell receptive field properties by learning a sparse code for natural images
TLDR
It is shown that a learning algorithm that attempts to find sparse linear codes for natural scenes will develop a complete family of localized, oriented, bandpass receptive fields, similar to those found in the primary visual cortex. Expand
Deep Unsupervised Learning on a Desktop PC: A Primer for Cognitive Scientists
TLDR
It is shown how simulations of deep unsupervised learning can be easily performed on a desktop PC by exploiting the processors of low cost graphic cards without any specific programing effort, thanks to the use of high-level programming routines (available in MATLAB or Python). Expand
Modeling language and cognition with deep unsupervised learning: a tutorial overview
TLDR
It is argued that the focus on deep architectures and generative (rather than discriminative) learning represents a crucial step forward for the connectionist modeling enterprise, because it offers a more plausible model of cortical learning as well as a way to bridge the gap between emergentist connectionist models and structured Bayesian models of cognition. Expand
Learning Orthographic Structure With Sequential Generative Neural Networks
TLDR
This work investigates a sequential version of the restricted Boltzmann machine (RBM), a stochastic recurrent neural network that extracts high-order structure from sensory data through unsupervised generative learning and can encode contextual information in the form of internal, distributed representations. Expand
Deep Learning of Representations for Unsupervised and Transfer Learning
  • Yoshua Bengio
  • Computer Science
  • ICML Unsupervised and Transfer Learning
  • 2012
TLDR
Why unsupervised pre-training of representations can be useful, and how it can be exploited in the transfer learning scenario, where the authors care about predictions on examples that are not from the same distribution as the training distribution. Expand
Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects.
We describe a model of visual processing in which feedback connections from a higher- to a lower- order visual cortical area carry predictions of lower-level neural activities, whereas theExpand
The “independent components” of natural scenes are edge filters
TLDR
It is shown that a new unsupervised learning algorithm based on information maximization, a nonlinear "infomax" network, when applied to an ensemble of natural scenes produces sets of visual filters that are localized and oriented. Expand
A Simple Common Contexts Explanation for the Development of Abstract Letter Identities
TLDR
A self-organizing artificial neural network is presented that illustrates this idea and produces ALIs when presented with the most frequent words from a beginning reading corpus, as well as with artificial input. Expand
...
1
2
3
4
5
...