Share This Author
Discourse Embellishment Using a Deep Encoder-Decoder Network
A new NLG task, textual embellishment, is defined by taking a text as input and generating a semantically equivalent output with increased lexical and syntactic complexity, and is presented using LSTM Encoder-Decoder networks trained on the WikiLarge dataset.
Cerebral microbleed detection in traumatic brain injury patients using 3D convolutional neural networks
A computer aided detection system based on 3D convolutional neural networks for detecting CMBs in 3D susceptibility weighted images outperforming related work on CMB detection in TBI patients.
Visual Attention Through Uncertainty Minimization in Recurrent Generative Models
A recurrent generative neural network model is proposed that predicts a visual scene based on foveated glimpses and shifts its attention in order to minimize the uncertainty in its predictions, providing evidence that uncertainty minimization could be a fundamental mechanisms for the allocation of visual attention.
Encoding and Decoding Dynamic Sensory Signals with Recurrent Neural Networks: An Application of Conceptors to Birdsongs
- Richard Gast, Patrick Faion, K. Standvoss, Andrea Suckro, B. Lewis, G. Pipa
- Computer Science, BiologybioRxiv
- 28 April 2017
This work proposes a model that describes a way to encode and decode sensory stimuli based on the activity patterns of multiple, recurrently connected neural populations with different receptive fields and demonstrates the ability of this model to learn and recognize complex, dynamic stimuli using birdsongs as exemplary data.
Humans and mice fluctuate between external and internal modes of sensory processing
It is found that humans and mice waver between alternating intervals of externally- and internally-oriented modes of sensory analysis, and between-mode fluctuations may benefit perception by enabling the generation of stable representations of the environment despite an ongoing stream of noisy sensory inputs.
Bimodal inference in humans and mice
It is found that humans and mice waver between alternating intervals of externally- and internally-oriented modes of sensory analysis, and simulated data suggested that between-mode fluctuations may benefit perception by generating unambiguous error signals that enable robust learning and metacognition in volatile environments.
Uncertainty through Sampling: The Correspondence of Monte Carlo Dropout and Spiking in Artificial Neural Networks
It is demonstrated that in cases of incomplete world knowledge (epistemic uncertainty) as well as for noisy observations (aleatoric uncertainty) both neuron models show similar uncertainty representations, providing evidence that sampling could play a fundamental role in representing uncertainties in neural systems.
Task-Dependent Attention Allocation through Uncertainty Minimization in Deep Recurrent Generative Models
- K. Standvoss, Silvan C. Quax, M. V. van Gerven
- Psychology, Computer ScienceConference on Cognitive Computational…
A recurrent generative neural network model is proposed that predicts a visual scene based on foveated glimpses and produces naturalistic eye-movements focusing on salient stimulus regions, providing evidence that uncertainty minimization could be a fundamental mechanisms for the allocation of visual attention.