We introduce a method that takes advantage of high-quality pretrained multimodal representations to explore fine-grained semantic networks in the human brain. Previous studies have documented evidence of functional localization in the brain, with different anatomical regions preferentially activating for different types of sensory input. Many such localized structures are known, including the fusiform face area and parahippocampal place area. This raises the question of whether additional brain regions (or conjunctions of brain regions) are also specialized for other important semantic concepts. To identify such brain regions, we developed a data-driven approach to uncover visual concepts that are decodable from a massive functional magnetic resonance imaging (fMRI) dataset. Our analysis is broadly split into three sections. First, a fully connected neural network is trained to map brain responses to the outputs of an image-language foundation model, CLIP (Radford et al., 2021). Subsequently, a contrastive-learning dimensionality reduction method reveals the brain-decodable components of CLIP space. In the final section of our analysis, we localize shared decodable concepts in the brain using a voxel-masking optimization method to produce a shared decodable concept (SDC) space. The accuracy of our procedure is validated by comparing it to previous localization experiments that identify regions for faces, bodies, and places. In addition to these concepts, whose corresponding brain regions were already known, we localize novel concept representations which are shared across participants to other areas of the human brain. We also demonstrate how this method can be used to inspect fine-grained semantic networks for individual participants. We envisage that this extensible method can also be adapted to explore other questions at the intersection of AI and neuroscience.
As convolutional neural networks (CNNs) become more accurate at object recognition, their representations become more similar to the primate visual system. This finding has inspired us and other researchers to ask if the implication also runs the other way: If CNN representations become more brain-like, does the network become more accurate? Previous attempts to address this question showed very modest gains in accuracy, owing in part to limitations of the regularization method. To overcome these limitations, we developed a new neural data regularizer for CNNs that uses Deep Canonical Correlation Analysis (DCCA) to optimize the resemblance of the CNN's image representations to that of the monkey visual cortex. Using this new neural data regularizer, we see much larger performance gains in both classification accuracy and within-super-class accuracy, as compared to the previous state-of-the-art neural data regularizers. These networks are also more robust to adversarial attacks than their unregularized counterparts. Together, these results confirm that neural data regularization can push CNN performance higher, and introduces a new method that obtains a larger performance boost.
Recent studies suggest that artificial neural networks (ANNs) that match the spectral properties of the mammalian visual cortex -- namely, the $\sim 1/n$ eigenspectrum of the covariance matrix of neural activities -- achieve higher object recognition performance and robustness to adversarial attacks than those that do not. To our knowledge, however, no previous work systematically explored how modifying the ANN's spectral properties affects performance. To fill this gap, we performed a systematic search over spectral regularizers, forcing the ANN's eigenspectrum to follow $1/n^\alpha$ power laws with different exponents $\alpha$. We found that larger powers (around 2--3) lead to better validation accuracy and more robustness to adversarial attacks on dense networks. This surprising finding applied to both shallow and deep networks and it overturns the notion that the brain-like spectrum (corresponding to $\alpha \sim 1$) always optimizes ANN performance and/or robustness. For convolutional networks, the best $\alpha$ values depend on the task complexity and evaluation metric: lower $\alpha$ values optimized validation accuracy and robustness to adversarial attack for networks performing a simple object recognition task (categorizing MNIST images of handwritten digits); for a more complex task (categorizing CIFAR-10 natural images), we found that lower $\alpha$ values optimized validation accuracy whereas higher $\alpha$ values optimized adversarial robustness. These results have two main implications. First, they cast doubt on the notion that brain-like spectral properties ($\alpha \sim 1$) \emph{always} optimize ANN performance. Second, they demonstrate the potential for fine-tuned spectral regularizers to optimize a chosen design metric, i.e., accuracy and/or robustness.
The current state-of-the-art object recognition algorithms, deep convolutional neural networks (DCNNs), are inspired by the architecture of the mammalian visual system [8], and capable of human-level performance on many tasks [15]. However, even these algorithms make errors. As DCNNs improve at object recognition tasks, they develop representations in their hidden layers that become more similar to those observed in the mammalian brains [24]. This led us to hypothesize that teaching DCNNs to achieve even more brain-like representations could improve their performance. To test this, we trained DCNNs on a composite task, wherein networks were trained to: a) classify images of objects; while b) having intermediate representations that resemble those observed in neural recordings from monkey visual cortex. Compared with DCNNs trained purely for object categorization, DCNNs trained on the composite task had better object recognition performance. Our results outline a new way to regularize object recognition networks, using transfer learning strategies in which the brain serves as a teacher for training DCNNs.
Primary visual cortex (V1) is the first stage of cortical image processing, and a major effort in systems neuroscience is devoted to understanding how it encodes information about visual stimuli. Within V1, many neurons respond selectively to edges of a given preferred orientation: these are known as simple or complex cells, and they are well-studied. Other neurons respond to localized center-surround image features. Still others respond selectively to certain image stimuli, but the specific features that excite them are unknown. Moreover, even for the simple and complex cells-- the best-understood V1 neurons-- it is challenging to predict how they will respond to natural image stimuli. Thus, there are important gaps in our understanding of how V1 encodes images. To fill this gap, we train deep convolutional neural networks to predict the firing rates of V1 neurons in response to natural image stimuli, and find that 15% of these neurons are within 10% of their theoretical limit of predictability. For these well predicted neurons, we invert the predictor network to identify the image features (receptive fields) that cause the V1 neurons to spike. In addition to those with previously-characterized receptive fields (Gabor wavelet and center-surround), we identify neurons that respond predictably to higher-level textural image features that are not localized to any particular region of the image.