A5048551460- Memory and Neural Mechanisms
- Neuroscience and Neuropharmacology Research
- Neural dynamics and brain function
- Neural and Behavioral Psychology Studies
- Memory Processes and Influences
- Functional Brain Connectivity Studies
- Reinforcement Learning in Robotics
- Multimodal Machine Learning Applications
- Child and Animal Learning Development
- Domain Adaptation and Few-Shot Learning
- Decision-Making and Behavioral Economics
- Neural Networks and Applications
- Olfactory and Sensory Function Studies
- Treatment of Major Depression
- Robotic Path Planning Algorithms
- Sleep and Wakefulness Research
- Stress Responses and Cortisol
- Neural Networks and Reservoir Computing
- Neuroinflammation and Neurodegeneration Mechanisms
- Topic Modeling
- Explainable Artificial Intelligence (XAI)
- Complex Systems and Decision Making
- Neuroscience and Music Perception
- Natural Language Processing Techniques
- Face Recognition and Perception
Google (United Kingdom)
2016-2024
DeepMind (United Kingdom)
2016-2024
University College London
2011-2020
National Hospital for Neurology and Neurosurgery
2007-2018
Google (United States)
2015-2018
Imperial College London
2018
Stanford University
2009-2014
University of Delhi
2008-2011
Ambedkar University Delhi
2008-2011
Palo Alto University
2011
The ability to learn tasks in a sequential fashion is crucial the development of artificial intelligence. Until now neural networks have not been capable this and it has widely thought that catastrophic forgetting an inevitable feature connectionist models. We show possible overcome limitation train can maintain expertise on they experienced for long time. Our approach remembers old by selectively slowing down learning weights important those tasks. demonstrate our scalable effective solving...
The game of chess is the longest-studied domain in history artificial intelligence. strongest programs are based on a combination sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. By contrast, AlphaGo Zero program recently achieved superhuman performance Go reinforcement learning from self-play. In this paper, we generalize approach into single AlphaZero algorithm can achieve many...
Human choices are remarkably susceptible to the manner in which options presented. This so-called "framing effect" represents a striking violation of standard economic accounts human rationality, although its underlying neurobiology is not understood. We found that framing effect was specifically associated with amygdala activity, suggesting key role for an emotional system mediating decision biases. Moreover, across individuals, orbital and medial prefrontal cortex activity predicted...
Amnesic patients have a well established deficit in remembering their past experiences. Surprisingly, however, the question as to whether such can imagine new experiences has not been formally addressed our knowledge. We tested group of amnesic with primary damage hippocampus bilaterally could construct imagined response short verbal cues that outlined range simple commonplace scenarios. Our results revealed were markedly impaired relative matched control subjects at imagining Moreover, we...
The game of chess is the most widely-studied domain in history artificial intelligence. strongest programs are based on a combination sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. In contrast, AlphaGo Zero program recently achieved superhuman performance Go, tabula rasa reinforcement learning from games self-play. this paper, we generalise approach into single AlphaZero...
Functional MRI (fMRI) studies investigating the neural basis of episodic memory recall, and related task thinking about plausible personal future events, have revealed a consistent network associated brain regions. Surprisingly little, however, is understood contributions individual areas make to overall recollective experience. To examine this, we used novel fMRI paradigm in which subjects had imagine fictitious experiences. In contrast thinking, this results experiences that are not...
In this article, we present a perspective on the role of hippocampal system in generalization, instantiated computational model called REMERGE (recurrency and episodic memory results generalization).We expose fundamental, but neglected, tension between prevailing theories that emphasize function hippocampus pattern separation (Marr, 1971;McClelland, McNaughton, & O'Reilly, 1995), empirical support for its generalization flexible relational (Cohen Eichenbaum, 1993;Eichenbaum, 1999).Our...
The ability to identify and react novelty within the environment is fundamental survival. Computational models emphasize potential role of hippocampus in detection, its unique anatomical circuitry making it ideally suited act as a comparator between past present experience. hippocampus, therefore, viewed detect associative mismatches what expected based on retrieval experience current sensory input. However, direct evidence that human performs such operations lacking. We explored brain...
Learning to navigate in complex environments with dynamic elements is an important milestone developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. particular consider jointly goal-driven depth prediction loop closure classification tasks. This approach learn from raw input complicated...
In recent years deep reinforcement learning (RL) systems have attained superhuman performance in a number of challenging task domains. However, major limitation such applications is their demand for massive amounts training data. A critical present objective thus to develop RL methods that can adapt rapidly new tasks. the work we introduce novel approach this challenge, which refer as meta-reinforcement learning. Previous has shown recurrent networks support meta-learning fully supervised...
Concepts lie at the very heart of intelligence, providing organizing principles with which to comprehend world. Surprisingly little, however, is understood about how we acquire and deploy concepts. Here, show that a functionally coupled circuit involving hippocampus ventromedial prefrontal cortex (vMPFC) underpins emergence conceptual knowledge its effect on choice behavior. Critically, alone supported efficient transfer perceptually novel setting. These findings provide compelling evidence...
The hippocampus is proposed to be critical in distinguishing between similar experiences by performing pattern separation computations that create orthogonalized representations for related episodes. Previous neuroimaging studies have provided indirect evidence the dentate gyrus (DG) and CA3 hippocampal subregions support inferring nature of underlying from observation novelty signals. Here, we use ultra-high-resolution fMRI at 7 T multivariate analysis provide compelling DG subregion...
The hippocampus has long been proposed to play a critical role in novelty detection through its ability act as comparator between past and present experience. A recent study provided evidence for this hypothesis by characterizing hippocampal responses sequence novelty, type of associative where familiar items appear new temporal order. Here, we ask whether match-mismatch (i.e., comparator) mechanism operates selectively identify the violation predictions within domain or instead also...
The hippocampus is widely accepted to play a pivotal role in memory. Two influential theories offer competing accounts of its fundamental operating mechanism. cognitive map theory posits special mapping large-scale space, whereas the relational argues it supports amodal processing. Here, we pit two against each other using novel paradigm which processing involved navigating city was matched with similar navigational and demands nonspatial (social) domain. During functional magnetic resonance...
Genetic variation at the serotonin transporter-linked polymorphic region (5-HTTLPR) is associated with altered amygdala reactivity and lack of prefrontal regulatory control. Similar regions mediate decision-making biases driven by contextual cues ambiguity, for example “framing effect.” We hypothesized that individuals hemozygous short (s) allele 5-HTTLPR would be more susceptible to framing. Participants, selected as homozygous either long (la) or s allele, performed a task where they made...