Models of Acetylcholine and Dopamine Signals Differentially Improve Neural Representations

Reducing Catastrophic Forgetting With Associative Learning: A Lesson From Fruit Flies

Catastrophic forgetting remains an outstanding challenge in continual learning. Recently, methods inspired by the brain, such as continual representation learning and memory replay, have been used to combat catastrophic forgetting. Associative learning (retaining associations between inputs and outputs, even after good representations are learned) plays an important function in the brain; however, its role in continual learning has not been carefully studied. Here, we identified a two-layer neural circuit in the fruit fly olfactory system that performs continual associative learning between odors and their associated valences. In the first layer, inputs (odors) are encoded using sparse, high-dimensional representations, which reduces memory interference by activating nonoverlapping populations of neurons for different odors. In the second layer, only the synapses between odor-activated neurons and the odor's associated output neuron are modified during learning; the rest of the weights are frozen to prevent unrelated memories from being overwritten. We prove theoretically that these two perceptron-like layers help reduce catastrophic forgetting compared to the original perceptron algorithm, under continual learning. We then show empirically on benchmark data sets that this simple and lightweight architecture outperforms other popular neural-inspired algorithms when also using a two-layer feedforward architecture. Overall, fruit flies evolved an efficient continual associative learning algorithm, and circuit mechanisms from neuroscience can be translated to improve machine computation.

Neural Simpletrons: Learning in the Limit of Few Labels with Directed Generative Networks

We explore classifier training for data sets with very few labels. We investigate this task using a neural network for nonnegative data. The network is derived from a hierarchical normalized Poisson mixture model with one observed and two hidden layers. With the single objective of likelihood optimization, both labeled and unlabeled data are naturally incorporated into learning. The neural activation and learning equations resulting from our derivation are concise and local. As a consequence, the network can be scaled using standard deep learning tools for parallelized GPU implementation. Using standard benchmarks for nonnegative data, such as text document representations, MNIST, and NIST SD19, we study the classification performance when very few labels are used for training. In different settings, the network's performance is compared to standard and recently suggested semisupervised classifiers. While other recent approaches are more competitive for many labels or fully labeled data sets, we find that the network studied here can be applied to numbers of few labels where no other system has been reported to operate so far.

Optimal neural inference of stimulus intensities

In natural data, the class and intensity of stimuli are correlated. Current machine learning algorithms ignore this ubiquitous statistical property of stimuli, usually by requiring normalized inputs. From a biological perspective, it remains unclear how neural circuits may account for these dependencies in inference and learning. Here, we use a probabilistic framework to model class-specific intensity variations, and we derive approximate inference and online learning rules which reflect common hallmarks of neural computation. Concretely, we show that a neural circuit equipped with specific forms of synaptic and intrinsic plasticity (IP) can learn the class-specific features and intensities of stimuli simultaneously. Our model provides a normative interpretation of IP as a critical part of sensory learning and predicts that neurons can represent nontrivial input statistics in their excitabilities. Computationally, our approach yields improved statistical representations for realistic datasets in the visual and auditory domains. In particular, we demonstrate the utility of the model in estimating the contrastive stress of speech.

Increasing Resilience to Traumatic Stress: Understanding the Protective Role of Well-Being

The brain maintains homeostasis in part through a network of feedback and feed-forward mechanisms, where neurochemicals and immune markers act as mediators. Using a previously constructed model of biobehavioral feedback, we found that in addition to healthy equilibrium another stable regulatory program supported chronic depression and anxiety. Exploring mechanisms that might underlie the contributions of subjective well-being to improved therapeutic outcomes in depression, we iteratively screened 288 candidate feedback patterns linking well-being to molecular signaling networks for those that maintained the original homeostatic regimes. Simulating stressful trigger events on each candidate network while maintaining high levels of subjective well-being isolated a specific feedback network where well-being was promoted by dopamine and acetylcholine, and itself promoted norepinephrine while inhibiting cortisol expression. This biobehavioral feedback mechanism was especially effective in reproducing well-being's clinically documented ability to promote resilience and protect against onset of depression and anxiety.

The Insula and Taste Learning

The sense of taste is a key component of the sensory machinery, enabling the evaluation of both the safety as well as forming associations regarding the nutritional value of ingestible substances. Indicative of the salience of the modality, taste conditioning can be achieved in rodents upon a single pairing of a tastant with a chemical stimulus inducing malaise. This robust associative learning paradigm has been heavily linked with activity within the insular cortex (IC), among other regions, such as the amygdala and medial prefrontal cortex. A number of studies have demonstrated taste memory formation to be dependent on protein synthesis at the IC and to correlate with the induction of signaling cascades involved in synaptic plasticity. Taste learning has been shown to require the differential involvement of dopaminergic GABAergic, glutamatergic, muscarinic neurotransmission across an extended taste learning circuit. The subsequent activation of downstream protein kinases (ERK, CaMKII), transcription factors (CREB, Elk-1) and immediate early genes (c-fos, Arc), has been implicated in the regulation of the different phases of taste learning. This review discusses the relevant neurotransmission, molecular signaling pathways and genetic markers involved in novel and aversive taste learning, with a particular focus on the IC. Imaging and other studies in humans have implicated the IC in the pathophysiology of a number of cognitive disorders. We conclude that the IC participates in circuit-wide computations that modulate the interception and encoding of sensory information, as well as the formation of subjective internal representations that control the expression of motivated behaviors.