Twitter | Pretraživanje | |
Tim Kietzmann
New preprint from the lab: "Individual differences among deep neural network models." Work with , , and Courtney Spoerer. below. 1/7
Deep neural networks (DNNs) excel at visual recognition tasks and are increasingly used as a modelling framework for neural computations in the primate brain. However, each DNN instance, just like...
bioRxiv bioRxiv @biorxivpreprint
Reply Retweet Označi sa "sviđa mi se" More
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
Deep neural networks have seen a surge in popularity in neuroscience and psychology, where they are used as a modelling framework to understand (visual) information processing in the brain. 2/7
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
A computationally convenient (and therefore common) approach is to rely on single pre-trained computer vision models (Alexnet, VGG, etc.). But do DNNs, just like brains, exhibit individual representational differences that need to be accounted for? 3/7
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
Here we test this by training multiple identical network instances while varying only the random seed during weight initialisation. We compare the learned representations using a technique from systems neuroscience: representational similarity analysis (RSA). 4/7
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
Simply changing the random seed leads to considerable individual differences (shared variance in distance estimates can be as low as 44% across networks). The size of the effect is comparable to training networks with completely different image sets. 5/7
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
What are the origins of this? We argue that the categorization objective does not sufficiently constrain the arrangement of category clusters and exemplars. In addition, the interplay of ReLus and properties of certain distance measures contribute to differences. 6/7
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Odgovor korisniku/ci @TimKietzmann
Dropout can help, but considerable differences remain. This calls into question the practice of using single network instances to derive neuroscientific insight. Going forward, multiple DNNs may need to be analysed (similar to experimental participants). /fin
Reply Retweet Označi sa "sviđa mi se"
Fleur Zeldenrust 10. sij
Interesting! So for me the question is: similarly, all 'real' brains are different. But what are the 'conserved' representations?
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
Conserved across human individuals, or between DNNs and brains?
Reply Retweet Označi sa "sviđa mi se"
Nicholas Blauch 10. sij
any thoughts on which constrains individual differences more: randomized initialization, or randomized training order?
Reply Retweet Označi sa "sviđa mi se"
Tim Kietzmann 10. sij
A good question to which I have no definite answer. We have compared differences that emerge from different random seeds (smallest intervention), differences due to different image sets (same categories), and differences due to different categories (Figure 5 in the paper).
Reply Retweet Označi sa "sviđa mi se"