Understanding how the brain learns throughout a lifetime remains a long-standing challenge. In artificial neural networks (ANNs), incorporating novel information too rapidly results in catastrophic interference, i.e., abrupt loss of previously acquired knowledge. Complementary Learning Systems Theory (CLST) suggests that new memories can be gradually integrated into the neocortex by interleaving new memories with existing knowledge. This approach, however, has been assumed to require interleaving all existing knowledge every time something new is learned, which is implausible because it is time-consuming and requires a large amount of data. We show that deep, nonlinear ANNs can learn new information by interleaving only a subset of old items that share substantial representational similarity with the new information. By using such similarity-weighted interleaved learning (SWIL), ANNs can learn new information rapidly with a similar accuracy level and minimal interference, while using a much smaller number of old items presented per epoch (fast and data-efficient). SWIL is shown to work with various standard classification datasets (Fashion-MNIST, CIFAR10, and CIFAR100), deep neural network architectures, and in sequential learning frameworks. We show that data efficiency and speedup in learning new items are increased roughly proportionally to the number of nonoverlapping classes stored in the network, which implies an enormous possible speedup in human brains, which encode a high number of separate categories. Finally, we propose a theoretical model of how SWIL might be implemented in the brain.

In this study, we explore the mechanistic relationship between growth factor signaling and kinase activity that supports the protein synthesis-dependent phase of long-term memory (LTM) consolidation for sensitization ofAplysia Specifically, we examine LTM for tail shock-induced sensitization of the tail-elicited siphon withdrawal (T-SW) reflex, a form of memory that requires both (i) extracellular signal-regulated kinase (ERK1/2; MAPK) activity within identified sensory neurons (SNs) that mediate the T-SW and (ii) the activation of transforming growth factor β (TGFβ) signaling. We now report that repeated tail shocks that induce intermediate-term (ITM) and LTM for sensitization, also induce a sustained post-training phase of MAPK activity in SNs (lasting at least 1 h). We identified two mechanistically distinct phases of post-training MAPK: (i) an immediate phase that does not require ongoing protein synthesis or TGFβ signaling, and (ii) a sustained phase that requires both protein synthesis and extracellular TGFβ signaling. We find that LTM consolidation requires sustained MAPK, and is disrupted by inhibitors of protein synthesis and TGFβ signaling during the consolidation window. These results provide strong evidence that TGFβ signaling sustains MAPK activity as an essential mechanistic step for LTM consolidation.

There is now compelling evidence that the allocation of memory to specific neurons (neuronal allocation) and synapses (synaptic allocation) in a neurocircuit is not random and that instead specific mechanisms, such as increases in neuronal excitability and synaptic tagging and capture, determine the exact sites where memories are stored. We propose an integrated view of these processes, such that neuronal allocation, synaptic tagging and capture, spine clustering and metaplasticity reflect related aspects of memory allocation mechanisms. Importantly, the properties of these mechanisms suggest a set of rules that profoundly affect how memories are stored and recalled.

The prevailing view is that striatal parvalbumin (PV)-positive interneurons primarily function to downregulate medium spiny projection neuron (MSN) activity via monosynaptic inhibitory signaling. Here, by combining in vivo neural recordings and optogenetics, we unexpectedly find that both suppressing and over-activating PV cells attenuates spontaneous MSN activity. To account for this, we find that, in addition to monosynaptic coupling, PV-MSN interactions are mediated by a competing disynaptic inhibitory circuit involving a variety of neuropeptide Y-expressing interneurons. Next we use optogenetic and chemogenetic approaches to show that dorsolateral striatal PV interneurons influence the initial expression of reward-conditioned responses but that their contribution to performance declines with experience. Consistent with this, we observe with large-scale recordings in behaving animals that the relative contribution of PV cells on MSN activity diminishes with training. Together, this work provides a possible mechanism by which PV interneurons modulate striatal output and selectively enhance performance early in learning.

The molecular and cellular mechanisms of memory storage have attracted a great deal of attention. By comparison, little is known about memory allocation, the process that determines which specific neurons in a neural network will store a given memory. Previous studies demonstrated that memory allocation is not random in the amygdala; these studies showed that amygdala neurons with higher levels of the cyclic-AMP-response-element-binding protein (CREB) are more likely to be recruited into encoding and storing fear memory. To determine whether specific mechanisms also regulate memory allocation in other brain regions and whether CREB also has a role in this process, we studied insular cortical memory representations for conditioned taste aversion (CTA). In this task, an animal learns to associate a taste (conditioned stimulus [CS]) with the experience of malaise (such as that induced by LiCl; unconditioned stimulus [US]). The insular cortex is required for CTA memory formation and retrieval. CTA learning activates a subpopulation of neurons in this structure, and the insular cortex and the basolateral amygdala (BLA) interact during CTA formation. Here, we used a combination of approaches, including viral vector transfections of insular cortex, arc fluorescence in situ hybridization (FISH), and designer receptors exclusively activated by designer drugs (DREADD) system, to show that CREB levels determine which insular cortical neurons go on to encode a given conditioned taste memory.

Extinction is the primary mode for the treatment of anxiety disorders. However, extinction memories are prone to relapse. For example, fear is likely to return when a prolonged time period intervenes between extinction and a subsequent encounter with the fear-provoking stimulus (spontaneous recovery). Therefore there is considerable interest in the development of procedures that strengthen extinction and to prevent such recovery of fear. We contrasted two procedures in rats that have been reported to cause such deepened extinction. One where extinction begins before the initial consolidation of fear memory begins (immediate extinction) and another where extinction begins after a brief exposure to the consolidated fear stimulus. The latter is thought to open a period of memory vulnerability similar to that which occurs during initial consolidation (reconsolidation update). We also included a standard extinction treatment and a control procedure that reversed the brief exposure and extinction phases. Spontaneous recovery was only found with the standard extinction treatment. In a separate experiment we tested fear shortly after extinction (i.e., within 6 h). All extinction procedures, except reconsolidation update reduced fear at this short-term test. The findings suggest that strengthened extinction can result from alteration in both retrieval and consolidation processes.

Recent studies suggest that a shared neural ensemble may link distinct memories encoded close in time. According to the memory allocation hypothesis, learning triggers a temporary increase in neuronal excitability that biases the representation of a subsequent memory to the neuronal ensemble encoding the first memory, such that recall of one memory increases the likelihood of recalling the other memory. Here we show in mice that the overlap between the hippocampal CA1 ensembles activated by two distinct contexts acquired within a day is higher than when they are separated by a week. Several findings indicate that this overlap of neuronal ensembles links two contextual memories. First, fear paired with one context is transferred to a neutral context when the two contexts are acquired within a day but not across a week. Second, the first memory strengthens the second memory within a day but not across a week. Older mice, known to have lower CA1 excitability, do not show the overlap between ensembles, the transfer of fear between contexts, or the strengthening of the second memory. Finally, in aged mice, increasing cellular excitability and activating a common ensemble of CA1 neurons during two distinct context exposures rescued the deficit in linking memories. Taken together, these findings demonstrate that contextual memories encoded close in time are linked by directing storage into overlapping ensembles. Alteration of these processes by ageing could affect the temporal structure of memories, thus impairing efficient recall of related information.