An official website of the United States government
According to the efficient coding hypothesis, neural populations encode information optimally when representations are high-dimensional and uncorrelated. However, such codes may carry a cost in terms of generalization and robustness. Past empirical studies of early visual cortex (V1) in rodents have suggested that this tradeoff indeed constrains sensory representations. However, it remains unclear whether these insights generalize across the hierarchy of the human visual system, and particularly to object representations in high-level occipitotemporal cortex (OTC). To gain new empirical clarity, here we develop a family of object recognition models with parametrically varying dropout proportion , which induces systematically varying dimensionality of internal responses (while controlling all other inductive biases). We find that increasing dropout produces an increasingly smooth, low-dimensional representational space. Optimal robustness to lesioning is observed at around 70% dropout, after which both accuracy and robustness decline. Representational comparison to large-scale 7T fMRI data from occipitotemporal cortex in the Natural Scenes Dataset reveals that this optimal degree of dropout is also associated with maximal emergent neural predictivity. Finally, using new techniques for achieving denoised estimates of the eigenspectrum of human fMRI responses, we compare the rate of eigenspectrum decay between model and brain feature spaces. We observe that the match between model and brain representations is associated with a common balance between efficiency and robustness in the representational space. These results suggest that varying dropout may reveal an optimal point of balance between the efficiency of high-dimensional codes and the robustness of low dimensional codes in hierarchical vision systems.
more »
« less
A case for sparse positive alignment of neural systems
Brain responses in visual cortex are typically modeled as a positively and negatively weighted sum of all features within a deep neural network (DNN) layer. However, this linear fit can dramatically alter a given feature space, making it unclear whether brain prediction levels stem more from the DNN itself, or from the flexibility of the encoding model. As such, studies of alignment may benefit from a paradigm shift toward more constrained and theoretically driven mapping methods. As a proof of concept, here we present a case study of face and scene selectivity, showing that typical encoding analyses do not differentiate between aligned and misaligned tuning bases in model-to-brain predictivity. We introduce a new alignment complexity measure -- tuning reorientation -- which favors DNNs that achieve high brain alignment via minimal distortion of the original feature space. We show that this measure helps arbitrate between models that are superficially equal in their predictivity, but which differ in alignment complexity. Our experiments broadly signal the benefit of sparse, positive-weighted encoding procedures, which directly enforce an analogy between the tuning directions of model and brain feature spaces.
more »
« less
- Award ID(s):
- 1942438
- PAR ID:
- 10510919
- Publisher / Repository:
- Open Review
- Date Published:
- Journal Name:
- ICLR 2024 Workshop Re-Align
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Discovering word-like units without textual transcriptions is an important step in low-resource speech technology. In this work,we demonstrate a model inspired by statistical machine translation and hidden Markov model/deep neural network (HMM-DNN) hybrid systems. Our learning algorithm is capable of discovering the visual and acoustic correlates of distinct words in an unknown language by simultaneously learning the map-ping from image regions to concepts (the first DNN), the map-ping from acoustic feature vectors to phones (the second DNN),and the optimum alignment between the two (the HMM). In the simulated low-resource setting using MSCOCO and Speech-COCO datasets, our model achieves 62.4 % alignment accuracy and outperforms the audio-only segmental embedded GMM approach on standard word discovery evaluation metrics.more » « less
-
The modeling of the brain as a three-dimensional spatial object, similar to a geographical landscape, has the paved way for the successful application of Kriging methods in solving the seizure detection problem with good performance but in cubic computational time complexity. The Deep Neural Network (DNN) has been widely used for seizure detection due to its effectiveness in classification tasks, although at the cost of a protracted training time. While Kriging exploits the spatial correlation between data locations, DNN relies on its capacity to learn intrinsic representations within the dataset from the basest unit parts. This paper presents a Distributed Kriging-Bootstrapped Deep Neural Network (DNN) model as a twofold solution for fast and accurate seizure detection using brain signals collected with the electroencephalogram (EEG) from healthy subjects and patients of epilepsy. The proposed model parallelizes the Kriging computation into different cores in a machine and then produces a strongly correlated, unified quasi-output data which serves as an input to the Deep Neural Network. Experimental results validate the proposed model as superior to conventional Kriging methods and DNN by training in 91% less time than the basic DNN and about three times as fast as the ordinary Kriging-Bootstrapped Deep Neural Network (DNN) model while maintaining good performance in terms of sensitivity, specificity and testing accuracy compared to other models and existing works.more » « less
-
Abstract This paper explicates a solution to building correspondences between molecular-scale transcriptomics and tissue-scale atlases. This problem arises in atlas construction and cross-specimen/technology alignment where specimens per emerging technology remain sparse and conventional image representations cannot efficiently model the high dimensions from subcellular detection of thousands of genes. We address these challenges by representing spatial transcriptomics data as generalized functions encoding position and high-dimensional feature (gene, cell type) identity. We map onto low-dimensional atlas ontologies by modeling regions as homogeneous random fields with unknown transcriptomic feature distribution. We solve simultaneously for the minimizing geodesic diffeomorphism of coordinates through LDDMM and for these latent feature densities. We map tissue-scale mouse brain atlases to gene-based and cell-based transcriptomics data from MERFISH and BARseq technologies and to histopathology and cross-species atlases to illustrate integration of diverse molecular and cellular datasets into a single coordinate system as a means of comparison and further atlas construction.more » « less
-
Testing deep neural networks (DNNs) has garnered great interest in the recent years due to their use in many applications. Black-box test adequacy measures are useful for guiding the testing process in covering the input domain. However, the absence of input specifications makes it challenging to apply black-box test adequacy measures in DNN testing. The Input Distribution Coverage (IDC) framework addresses this challenge by using a variational autoencoder to learn a low dimensional latent representation of the input distribution, and then using that latent space as a coverage domain for testing. IDC applies combinatorial interaction testing on a partitioning of the latent space to measure test adequacy. Empirical evaluation demonstrates that IDC is cost-effective, capable of detecting feature diversity in test inputs, and more sensitive than prior work to test inputs generated using different DNN test generation methods. The findings demonstrate that IDC overcomes several limitations of white-box DNN coverage approaches by discounting coverage from unrealistic inputs and enabling the calculation of test adequacy metrics that capture the feature diversity present in the input space of DNNs.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
