Neuroscience-Inspired Artificial Intelligence

Neuron, Год журнала: 2017, Номер 95(2), С. 245 - 258

Опубликована: Июль 1, 2017

Язык: Английский

---

Neuroscience Needs Behavior: Correcting a Reductionist Bias

Neuron, Год журнала: 2017, Номер 93(3), С. 480 - 490

Опубликована: Фев. 1, 2017

Язык: Английский

---

Toward an Integration of Deep Learning and Neuroscience

Frontiers in Computational Neuroscience, Год журнала: 2016, Номер 10

Опубликована: Сен. 14, 2016

Neuroscience has focused on the detailed implementation of computation, studying neural codes, dynamics and circuits. In machine learning, however, artificial networks tend to eschew precisely designed or circuits in favor brute force optimization a cost function, often using simple relatively uniform initial architectures. Two recent developments have emerged within learning that create an opportunity connect these seemingly divergent perspectives. First, structured architectures are used, including dedicated systems for attention, recursion various forms short- long-term memory storage. Second, functions training procedures become more complex varied across layers over time. Here we think about brain terms ideas. We hypothesize (1) optimizes functions, (2) diverse differ locations development, (3) operates pre-structured architecture matched computational problems posed by behavior. Such heterogeneously optimized system, enabled series interacting serves make data-efficient targeted needs organism. suggest directions which neuroscience could seek refine test hypotheses.

Язык: Английский

---

Thalamic functions in distributed cognitive control

Nature Neuroscience, Год журнала: 2017, Номер 20(12), С. 1669 - 1679

Опубликована: Ноя. 9, 2017

Язык: Английский

---

Neural Entrainment and Attentional Selection in the Listening Brain

Trends in Cognitive Sciences, Год журнала: 2019, Номер 23(11), С. 913 - 926

Опубликована: Окт. 9, 2019

Язык: Английский

---

Direct Fit to Nature: An Evolutionary Perspective on Biological and Artificial Neural Networks

Neuron, Год журнала: 2020, Номер 105(3), С. 416 - 434

Опубликована: Фев. 1, 2020

Язык: Английский

---

Representational models: A common framework for understanding encoding, pattern-component, and representational-similarity analysis

PLoS Computational Biology, Год журнала: 2017, Номер 13(4), С. e1005508 - e1005508

Опубликована: Апрель 24, 2017

Representational models specify how activity patterns in populations of neurons (or, more generally, multivariate brain-activity measurements) relate to sensory stimuli, motor responses, or cognitive processes. In an experimental context, representational can be defined as hypotheses about the distribution profiles across conditions. Currently, three different methods are being used test such hypotheses: encoding analysis, pattern component modeling (PCM), and similarity analysis (RSA). Here we develop a common mathematical framework for understanding relationship these methods, which share one core commonality: all evaluate second moment profiles, determines geometry, thus well any feature decoded from population activity. Using simulated data designs, compare power adjudicate between competing models. PCM implements likelihood-ratio therefore provides most powerful if its assumptions hold. However, other two approaches-when conducted appropriately-can perform similarly. linear model needs appropriately regularized, effectively imposes prior on profiles. With prior, specifies well-defined RSA, unequal variances statistical dependencies dissimilarity estimates need taken into account reach near-optimal inference. The render aspects information explicit (e.g. single-response tuning population-response RSA) have specific advantages terms computational demands, ease use, extensibility. properly construed complementary components single data-analytical toolkit neural representations basis data.

Язык: Английский

---

Lack of Theory Building and Testing Impedes Progress in The Factor and Network Literature

Psychological Inquiry, Год журнала: 2020, Номер 31(4), С. 271 - 288

Опубликована: Окт. 1, 2020

The applied social science literature using factor and network models continues to grow rapidly. Most work reads like an exercise in model fitting, falls short of theory building testing three ways. First, statistical theoretical are conflated, leading invalid inferences such as the existence psychological constructs based on models, or recommendations for clinical interventions models. I demonstrate this inferential gap a simulation: excellent fit does little corroborate theory, regardless quality quantity data. Second, researchers fail explicate theories about constructs, but use implicit causal beliefs guide inferences. These latent have led problematic best practices. Third, explicated often weak theories: imprecise descriptions vulnerable hidden assumptions unknowns. Such do not offer precise predictions, it is unclear whether effects actually not. that these challenges common harmful, impede formation, failure, reform. Matching necessary bring data bear theories, renewed focus psychology formalizing offers way forward.

Язык: Английский

---

Causal mapping of human brain function

Nature reviews. Neuroscience, Год журнала: 2022, Номер 23(6), С. 361 - 375

Опубликована: Апрель 20, 2022

Язык: Английский

---

Are We Ready for Real-world Neuroscience?

Journal of Cognitive Neuroscience, Год журнала: 2018, Номер 31(3), С. 327 - 338

Опубликована: Июнь 19, 2018

Real-world environments are typically dynamic, complex, and multisensory in nature require the support of top-down attention memory mechanisms for us to be able drive a car, make shopping list, or pour cup coffee. Fundamental principles perception functional brain organization have been established by research utilizing well-controlled but simplified paradigms with basic stimuli. The last 30 years ushered revolution computational power, mapping, signal processing techniques. Drawing on those theoretical methodological advances, over years, has departed more from traditional, rigorous, well-understood directly investigate cognitive functions their underlying real-world environments. These investigations address role one or, recently, multiple attributes assumptions about perception, attention, challenged-by studies adapting traditional emulate, example, varying relevance stimulation dynamically changing task demands. Here, we present state field within emerging heterogeneous domain neuroscience. To precise, aim this Special Focus is bring together variety "real-world neuroscientific" approaches. approaches differ principal aims, assumptions, even definitions neuroscience" research. showcase commonalities distinctive features different do so, four early-career researchers speakers Cognitive Neuroscience Society 2017 Meeting symposium under same title answer questions pertaining added value such bringing closer accurate models functions.

Язык: Английский

---