Many recent studies find striking similarities between representations in biological brains
and artificial neural networks
trained to solve analogous tasks.
-
-
Prikaži ovu nit
-
This is pretty crazy when you think about it because brains and ANNs have serious differences in their biophysical/architectural details.
Prikaži ovu nit -
This raises a fundamental question: How should we interpret the often striking representational similarity of biological and artificial networks?

Prikaži ovu nit -
While the technology is not (yet!) there to address this question with biological brains, we can begin to address it theoretically by comparing various ANN architectures’ representations and dynamics against each other.
Prikaži ovu nit -
E.g. when you train a recurrent network on the same task with different architectures, would you expect the solutions to look different or the same? In what ways?
Prikaži ovu nit -
We trained thousands of RNNs to solve simple tasks, to see whether these comparisons (and scientific conclusions) would be sensitive to particularities of modeling choices, e.g. LSTM vs. GRU vs vanilla RNN.
Prikaži ovu nit -
We found evidence for both individuality
and universality
in the solutions across different RNN architectures, with the geometry of representations tending to be more varied and the dynamics tending to be more universal.Prikaži ovu nit -
Here’s an example from the preprint, a simple task to produce a sine wave whose frequency is proportional to a given input (command) frequency. We trained recurrent networks to solve this taskpic.twitter.com/AVebkXp8Kr
Prikaži ovu nit -
Example state space trajectories show some differences across architecturepic.twitter.com/RNc3ePYb3b
Prikaži ovu nit -
We compared the geometry of network representations using canonical correlation analysis (CCA). CCA suggests that networks are sensitive to these modeling choices (each dot is a network) if intracluster distances gives a measure for intercluster distance.pic.twitter.com/nMG5LES3r1
Prikaži ovu nit -
Next, we asked if we would reach the same conclusion if we compared the underlying dynamics
. To do this, we used tools from dynamical systems theory (fixed points and linearization) to extract a simple dynamical portrait for each network.pic.twitter.com/ARueBiVffQ
Prikaži ovu nit -
When we then compared network distances using fixed point topology, we found that there was no real difference across architecture (e), suggesting the topology may be universal.pic.twitter.com/yMCy4Ds1qe
Prikaži ovu nit -
Linearization of the dynamics reveals a common motif with some small differences across architectures. The common motif is a nearly linear solution to produce the oscillations. The varied motif is a small degree of nonlinearity used to generate the oscillations.pic.twitter.com/Uj1a6GTLZz
Prikaži ovu nit -
We hope this kind of in silico
study helps advance a discussion about the use of ANNs in neuroscience. There are more examples in the preprint https://arxiv.org/abs/1907.08549 .Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.


