1 code implementation • 14 Sep 2023 • Gašper Beguš, Thomas Lu, Alan Zhou, Peter Wu, Gopala K. Anumanchipalli
This paper introduces CiwaGAN, a model of human spoken language acquisition that combines unsupervised articulatory modeling with an unsupervised model of information exchange through the auditory modality.
no code implementations • 12 Jun 2023 • Maksymilian Dąbkowski, Gašper Beguš
Recursion is one of the hallmarks of human language.
no code implementations • 2 May 2023 • Gašper Beguš, Thomas Lu, Zili Wang
Computational models of syntax are predominantly text-based.
no code implementations • 1 May 2023 • Gašper Beguš, Maksymilian Dąbkowski, Ryan Rhodes
We show here that for the first time, the models can also generate coherent and valid formal analyses of linguistic data and illustrate the vast potential of large language models for analyses of their metalinguistic abilities.
1 code implementation • 25 Apr 2023 • Russell A Poldrack, Thomas Lu, Gašper Beguš
We report several experiments using GPT-4 to generate computer code.
1 code implementation • 20 Mar 2023 • Gašper Beguš, Andrej Leban, Shane Gero
This paper suggests that an interpretation of the outputs of deep neural networks with causal inference methodology can be a viable strategy for approaching data about which little is known and presents another case of how deep learning can limit the hypothesis space.
1 code implementation • 27 Oct 2022 • Gašper Beguš, Alan Zhou, Peter Wu, Gopala K Anumanchipalli
Articulatory analysis suggests that the network learns to control articulators in a similar manner to humans during speech production.
no code implementations • 22 Mar 2022 • Gašper Beguš, Alan Zhou
Here, we test how encoding and decoding of lexical semantic information can emerge automatically from raw speech in unsupervised generative deep convolutional networks that combine the production and perception principles of speech.
no code implementations • 5 Oct 2021 • Gašper Beguš, Alan Zhou
We propose a technique to visualize individual convolutional layers in the classifier that yields highly informative time-series data for each convolutional layer and apply it to unobserved test data.
no code implementations • 19 Apr 2021 • Gašper Beguš, Alan Zhou
This technique allows for acoustic analysis of intermediate layers that parallels the acoustic analysis of human speech data: we can extract F0, intensity, duration, formants, and other acoustic properties from intermediate layers in order to test where and how CNNs encode various types of information.
no code implementations • 17 Apr 2021 • Jacob Andreas, Gašper Beguš, Michael M. Bronstein, Roee Diamant, Denley Delaney, Shane Gero, Shafi Goldwasser, David F. Gruber, Sarah de Haas, Peter Malkin, Roger Payne, Giovanni Petri, Daniela Rus, Pratyusha Sharma, Dan Tchernov, Pernille Tønnesen, Antonio Torralba, Daniel Vogt, Robert J. Wood
We posit that machine learning will be the cornerstone of future collection, processing, and analysis of multimodal streams of data in animal communication studies, including bioacoustic, behavioral, biological, and environmental data.
no code implementations • 10 Nov 2020 • Gašper Beguš
This paper proposes a framework for modeling sound change that combines deep learning and iterative learning.
no code implementations • 27 Sep 2020 • Gašper Beguš
The proposed technique for retrieving learning representations has general implications for our understanding of how GANs discretize continuous speech data and suggests that rule-like generalizations in the training data are represented as an interaction between variables in the network's latent space.
1 code implementation • 13 Sep 2020 • Gašper Beguš
This paper models unsupervised learning of an identity-based pattern (or copying) in speech called reduplication from raw continuous data with deep convolutional neural networks.
no code implementations • 6 Jun 2020 • Gašper Beguš
A Generative Adversarial Network was trained on an allophonic distribution in English.
1 code implementation • 4 Jun 2020 • Gašper Beguš
The networks trained on lexical items from TIMIT learn to encode unique information corresponding to lexical items in the form of categorical variables in their latent space.