A5050749858- Memory Processes and Influences
- Financial Markets and Investment Strategies
- Forecasting Techniques and Applications
- Neural Networks and Applications
- Decision-Making and Behavioral Economics
- Functional Brain Connectivity Studies
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Cognitive Functions and Memory
- AI-based Problem Solving and Planning
- Child and Animal Learning Development
- Microfluidic and Bio-sensing Technologies
- Innovative Microfluidic and Catalytic Techniques Innovation
- EEG and Brain-Computer Interfaces
- 3D Printing in Biomedical Research
University of Oxford
2018-2024
The University of Tokyo
2023
Significance Humans learn to perform many different tasks over the lifespan, such as speaking both French and Spanish. The brain has represent task information without mutual interference. In machine learning, this “continual learning” is a major unsolved challenge. Here, we studied patterns of errors made by humans state-of-the-art neural networks while they learned new from scratch instruction. Humans, but not machines, seem benefit training regimes that blocked one at time, especially...
Generalization (or transfer) is the ability to repurpose knowledge in novel settings. It often asserted that generalization an important ingredient of human intelligence, but its extent, nature, and determinants have proved controversial. Here, we examine this with a paradigm formalizes transfer learning problem as one recomposing existing functions solve unseen problems. We find people can generalize compositionally ways are elusive for standard neural networks benefits from training...
Generalisation (or transfer) is the ability to repurpose knowledge in novel settings. It often asserted that generalisation an important ingredient of human intelligence, but its extent, nature and determinants have proved controversial. Here, we re-examine this question with a new paradigm formalises transfer learning problem as one recomposing existing functions solve unseen problems. We find people can generalise compositionally ways are elusive for standard neural networks, benefits from...
Experiments on decision making under uncertainty are known to display a classical pattern of risk aversion and seeking referred as “fourfold pattern” (or “reflection effect”), but recent experiments varying the speed order mental processing have brought light more nuanced phenomenology. We model though Bayesian formalization anchor-and-adjust heuristic observed in empirical studies cognitive bias. Using only elementary assumptions constrained information processing, we able infer three...
Experiments on decision making under uncertainty are known to display a classical pattern of risk aversion and seeking referred as "fourfold pattern" (or "reflection effect") , but recent experiments varying the speed order mental processing have brought light more nuanced phenomenology. We model though Bayesian formalization anchor-and-adjust heuristic observed in empirical studies cognitive bias. Using only elementary assumptions constrained information processing, we able infer three...
Abstract Humans can learn to perform multiple tasks in succession over the lifespan (“continual” learning), whereas current machine learning systems fail. Here, we investigated cognitive mechanisms that permit successful continual humans. Unlike neural networks, humans were trained on temporally autocorrelated task objectives (focussed training) learned new more effectively, and performed better a later test involving randomly interleaved tasks. Analysis of error patterns suggested focussed...