skip to main content

Title: Reward-based training of recurrent neural networks for cognitive and value-based tasks
A major goal in neuroscience is to understand the relationship between an animal’s behavior and how this is encoded in the brain. Therefore, a typical experiment involves training an animal to perform a task and recording the activity of its neurons – brain cells – while the animal carries out the task. To complement these experimental results, researchers “train” artificial neural networks – simplified mathematical models of the brain that consist of simple neuron-like units – to simulate the same tasks on a computer. Unlike real brains, artificial neural networks provide complete access to the “neural circuits” responsible for a behavior, offering a way to study and manipulate the behavior in the circuit. One open issue about this approach has been the way in which the artificial networks are trained. In a process known as reinforcement learning, animals learn from rewards (such as juice) that they receive when they choose actions that lead to the successful completion of a task. By contrast, the artificial networks are explicitly told the correct action. In addition to differing from how animals learn, this limits the types of behavior that can be studied using artificial neural networks. Recent advances in the field of machine learning that combine reinforcement learning with artificial neural networks have now allowed Song et al. to train artificial networks to perform tasks in a way that mimics the way that animals learn. The networks consisted of two parts: a “decision network” that uses sensory information to select actions that lead to the greatest reward, and a “value network” that predicts how rewarding an action will be. Song et al. found that the resulting artificial “brain activity” closely resembled the activity found in the brains of animals, confirming that this method of training artificial neural networks may be a useful tool for neuroscientists who study the relationship between brains and behavior. The training method explored by Song et al. represents only one step forward in developing artificial neural networks that resemble the real brain. In particular, neural networks modify connections between units in a vastly different way to the methods used by biological brains to alter the connections between neurons. Future work will be needed to bridge this gap.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Memories are an important part of how we think, understand the world around us, and plan out future actions. In the brain, memories are thought to be stored in a region called the hippocampus. When memories are formed, neurons store events that occur around the same time together. This might explain why often, in the brains of animals, the activity associated with retrieving memories is not just a snapshot of what happened at a specific moment-- it can also include information about what the animal might experience next. This can have a clear utility if animals use memories to predict what they might experience next and plan out future actions. Mathematically, this notion of predictiveness can be summarized by an algorithm known as the successor representation. This algorithm describes what the activity of neurons in the hippocampus looks like when retrieving memories and making predictions based on them. However, even though the successor representation can computationally reproduce the activity seen in the hippocampus when it is making predictions, it is unclear what biological mechanisms underpin this computation in the brain. Fang et al. approached this problem by trying to build a model that could generate the same activity patterns computed by the successor representation using only biological mechanisms known to exist in the hippocampus. First, they used computational methods to design a network of neurons that had the biological properties of neural networks in the hippocampus. They then used the network to simulate neural activity. The results show that the activity of the network they designed was able to exactly match the successor representation. Additionally, the data resulting from the simulated activity in the network fitted experimental observations of hippocampal activity in Tufted Titmice. One advantage of the network designed by Fang et al. is that it can generate predictions in flexible ways,. That is, it canmake both short and long-term predictions from what an individual is experiencing at the moment. This flexibility means that the network can be used to simulate how the hippocampus learns in a variety of cognitive tasks. Additionally, the network is robust to different conditions. Given that the brain has to be able to store memories in many different situations, this is a promising indication that this network may be a reasonable model of how the brain learns. The results of Fang et al. lay the groundwork for connecting biological mechanisms in the hippocampus at the cellular level to cognitive effects, an essential step to understanding the hippocampus, as well as its role in health and disease. For instance, their network may provide a concrete approach to studying how disruptions to the ways neurons make and break connections can impair memory formation. More generally, better models of the biological mechanisms involved in making computations in the hippocampus can help scientists better understand and test out theories about how memories are formed and stored in the brain. 
    more » « less
  2. Understanding the intricacies of the brain often requires spotting and tracking specific neurons over time and across different individuals. For instance, scientists may need to precisely monitor the activity of one neuron even as the brain moves and deforms; or they may want to find universal patterns by comparing signals from the same neuron across different individuals. Both tasks require matching which neuron is which in different images and amongst a constellation of cells. This is theoretically possible in certain ‘model’ animals where every single neuron is known and carefully mapped out. Still, it remains challenging: neurons move relative to one another as the animal changes posture, and the position of a cell is also slightly different between individuals. Sophisticated computer algorithms are increasingly used to tackle this problem, but they are far too slow to track neural signals as real-time experiments unfold. To address this issue, Yu et al. designed a new algorithm based on the Transformer, an artificial neural network originally used to spot relationships between words in sentences. To learn relationships between neurons, the algorithm was fed hundreds of thousands of ‘semi-synthetic’ examples of constellations of neurons. Instead of painfully collated actual experimental data, these datasets were created by a simulator based on a few simple measurements. Testing the new algorithm on the tiny worm Caenorhabditis elegans revealed that it was faster and more accurate, finding corresponding neurons in about 10ms. The work by Yu et al. demonstrates the power of using simulations rather than experimental data to train artificial networks. The resulting algorithm can be used immediately to help study how the brain of C. elegans makes decisions or controls movements. Ultimately, this research could allow brain-machine interfaces to be developed. 
    more » « less
  3. INTRODUCTION A brainwide, synaptic-resolution connectivity map—a connectome—is essential for understanding how the brain generates behavior. However because of technological constraints imaging entire brains with electron microscopy (EM) and reconstructing circuits from such datasets has been challenging. To date, complete connectomes have been mapped for only three organisms, each with several hundred brain neurons: the nematode C. elegans , the larva of the sea squirt Ciona intestinalis , and of the marine annelid Platynereis dumerilii . Synapse-resolution circuit diagrams of larger brains, such as insects, fish, and mammals, have been approached by considering select subregions in isolation. However, neural computations span spatially dispersed but interconnected brain regions, and understanding any one computation requires the complete brain connectome with all its inputs and outputs. RATIONALE We therefore generated a connectome of an entire brain of a small insect, the larva of the fruit fly, Drosophila melanogaster. This animal displays a rich behavioral repertoire, including learning, value computation, and action selection, and shares homologous brain structures with adult Drosophila and larger insects. Powerful genetic tools are available for selective manipulation or recording of individual neuron types. In this tractable model system, hypotheses about the functional roles of specific neurons and circuit motifs revealed by the connectome can therefore be readily tested. RESULTS The complete synaptic-resolution connectome of the Drosophila larval brain comprises 3016 neurons and 548,000 synapses. We performed a detailed analysis of the brain circuit architecture, including connection and neuron types, network hubs, and circuit motifs. Most of the brain’s in-out hubs (73%) were postsynaptic to the learning center or presynaptic to the dopaminergic neurons that drive learning. We used graph spectral embedding to hierarchically cluster neurons based on synaptic connectivity into 93 neuron types, which were internally consistent based on other features, such as morphology and function. We developed an algorithm to track brainwide signal propagation across polysynaptic pathways and analyzed feedforward (from sensory to output) and feedback pathways, multisensory integration, and cross-hemisphere interactions. We found extensive multisensory integration throughout the brain and multiple interconnected pathways of varying depths from sensory neurons to output neurons forming a distributed processing network. The brain had a highly recurrent architecture, with 41% of neurons receiving long-range recurrent input. However, recurrence was not evenly distributed and was especially high in areas implicated in learning and action selection. Dopaminergic neurons that drive learning are amongst the most recurrent neurons in the brain. Many contralateral neurons, which projected across brain hemispheres, were in-out hubs and synapsed onto each other, facilitating extensive interhemispheric communication. We also analyzed interactions between the brain and nerve cord. We found that descending neurons targeted a small fraction of premotor elements that could play important roles in switching between locomotor states. A subset of descending neurons targeted low-order post-sensory interneurons likely modulating sensory processing. CONCLUSION The complete brain connectome of the Drosophila larva will be a lasting reference study, providing a basis for a multitude of theoretical and experimental studies of brain function. The approach and computational tools generated in this study will facilitate the analysis of future connectomes. Although the details of brain organization differ across the animal kingdom, many circuit architectures are conserved. As more brain connectomes of other organisms are mapped in the future, comparisons between them will reveal both common and therefore potentially optimal circuit architectures, as well as the idiosyncratic ones that underlie behavioral differences between organisms. Some of the architectural features observed in the Drosophila larval brain, including multilayer shortcuts and prominent nested recurrent loops, are found in state-of-the-art artificial neural networks, where they can compensate for a lack of network depth and support arbitrary, task-dependent computations. Such features could therefore increase the brain’s computational capacity, overcoming physiological constraints on the number of neurons. Future analysis of similarities and differences between brains and artificial neural networks may help in understanding brain computational principles and perhaps inspire new machine learning architectures. The connectome of the Drosophila larval brain. The morphologies of all brain neurons, reconstructed from a synapse-resolution EM volume, and the synaptic connectivity matrix of an entire brain. This connectivity information was used to hierarchically cluster all brains into 93 cell types, which were internally consistent based on morphology and known function. 
    more » « less
  4. Humans and most animals can learn new tasks without forgetting old ones. However, training artificial neural networks (ANNs) on new tasks typically causes them to forget previously learned tasks. This phenomenon is the result of “catastrophic forgetting,” in which training an ANN disrupts connection weights that were important for solving previous tasks, degrading task performance. Several recent studies have proposed methods to stabilize connection weights of ANNs that are deemed most important for solving a task, which helps alleviate catastrophic forgetting. Here, drawing inspiration from algorithms that are believed to be implemented in vivo, we propose a complementary method: adding a context-dependent gating signal, such that only sparse, mostly nonoverlapping patterns of units are active for any one task. This method is easy to implement, requires little computational overhead, and allows ANNs to maintain high performance across large numbers of sequentially presented tasks, particularly when combined with weight stabilization. We show that this method works for both feedforward and recurrent network architectures, trained using either supervised or reinforcement-based learning. This suggests that using multiple, complementary methods, akin to what is believed to occur in the brain, can be a highly effective strategy to support continual learning.

    more » « less
  5. INTRODUCTION Diverse phenotypes, including large brains relative to body size, group living, and vocal learning ability, have evolved multiple times throughout mammalian history. These shared phenotypes may have arisen repeatedly by means of common mechanisms discernible through genome comparisons. RATIONALE Protein-coding sequence differences have failed to fully explain the evolution of multiple mammalian phenotypes. This suggests that these phenotypes have evolved at least in part through changes in gene expression, meaning that their differences across species may be caused by differences in genome sequence at enhancer regions that control gene expression in specific tissues and cell types. Yet the enhancers involved in phenotype evolution are largely unknown. Sequence conservation–based approaches for identifying such enhancers are limited because enhancer activity can be conserved even when the individual nucleotides within the sequence are poorly conserved. This is due to an overwhelming number of cases where nucleotides turn over at a high rate, but a similar combination of transcription factor binding sites and other sequence features can be maintained across millions of years of evolution, allowing the function of the enhancer to be conserved in a particular cell type or tissue. Experimentally measuring the function of orthologous enhancers across dozens of species is currently infeasible, but new machine learning methods make it possible to make reliable sequence-based predictions of enhancer function across species in specific tissues and cell types. RESULTS To overcome the limits of studying individual nucleotides, we developed the Tissue-Aware Conservation Inference Toolkit (TACIT). Rather than measuring the extent to which individual nucleotides are conserved across a region, TACIT uses machine learning to test whether the function of a given part of the genome is likely to be conserved. More specifically, convolutional neural networks learn the tissue- or cell type–specific regulatory code connecting genome sequence to enhancer activity using candidate enhancers identified from only a few species. This approach allows us to accurately associate differences between species in tissue or cell type–specific enhancer activity with genome sequence differences at enhancer orthologs. We then connect these predictions of enhancer function to phenotypes across hundreds of mammals in a way that accounts for species’ phylogenetic relatedness. We applied TACIT to identify candidate enhancers from motor cortex and parvalbumin neuron open chromatin data that are associated with brain size relative to body size, solitary living, and vocal learning across 222 mammals. Our results include the identification of multiple candidate enhancers associated with brain size relative to body size, several of which are located in linear or three-dimensional proximity to genes whose protein-coding mutations have been implicated in microcephaly or macrocephaly in humans. We also identified candidate enhancers associated with the evolution of solitary living near a gene implicated in separation anxiety and other enhancers associated with the evolution of vocal learning ability. We obtained distinct results for bulk motor cortex and parvalbumin neurons, demonstrating the value in applying TACIT to both bulk tissue and specific minority cell type populations. To facilitate future analyses of our results and applications of TACIT, we released predicted enhancer activity of >400,000 candidate enhancers in each of 222 mammals and their associations with the phenotypes we investigated. CONCLUSION TACIT leverages predicted enhancer activity conservation rather than nucleotide-level conservation to connect genetic sequence differences between species to phenotypes across large numbers of mammals. TACIT can be applied to any phenotype with enhancer activity data available from at least a few species in a relevant tissue or cell type and a whole-genome alignment available across dozens of species with substantial phenotypic variation. Although we developed TACIT for transcriptional enhancers, it could also be applied to genomic regions involved in other components of gene regulation, such as promoters and splicing enhancers and silencers. As the number of sequenced genomes grows, machine learning approaches such as TACIT have the potential to help make sense of how conservation of, or changes in, subtle genome patterns can help explain phenotype evolution. Tissue-Aware Conservation Inference Toolkit (TACIT) associates genetic differences between species with phenotypes. TACIT works by generating open chromatin data from a few species in a tissue related to a phenotype, using the sequences underlying open and closed chromatin regions to train a machine learning model for predicting tissue-specific open chromatin and associating open chromatin predictions across dozens of mammals with the phenotype. [Species silhouettes are from PhyloPic] 
    more » « less