Parametric control of flexible timing through low-dimensional neural manifolds

Neuron, Journal Year: 2023, Volume and Issue: 111(5), P. 739 - 753.e8

Published: Jan. 13, 2023

Language: Английский

---

The neural bases for timing of durations

Nature reviews. Neuroscience, Journal Year: 2022, Volume and Issue: 23(11), P. 646 - 665

Published: Sept. 12, 2022

Language: Английский

---

How movements shape the perception of time

Trends in Cognitive Sciences, Journal Year: 2021, Volume and Issue: 25(11), P. 950 - 963

Published: Sept. 14, 2021

Language: Английский

---

Large-scale prediction of stream water quality using an interpretable deep learning approach

Journal of Environmental Management, Journal Year: 2023, Volume and Issue: 331, P. 117309 - 117309

Published: Jan. 17, 2023

Language: Английский

---

Amodal population clock in the primate medial premotor system for rhythmic tapping

Cell Reports, Journal Year: 2023, Volume and Issue: 42(10), P. 113234 - 113234

Published: Oct. 1, 2023

The neural substrate for beat extraction and response entrainment to rhythms is not fully understood. Here we analyze the activity of medial premotor neurons in monkeys performing isochronous tapping guided by brief flashing stimuli or auditory tones. population dynamics shared following properties across modalities: circular trajectories form a regenerating loop every produced interval; converge similar state space at times resetting clock; tempo synchronized encoded combination amplitude modulation temporal scaling. Notably, modality induces displacement visual subspaces without greatly altering time-keeping mechanism. These results suggest that interaction between cortex's amodal internal representation pulse modality-specific external input generates rhythmic clock whose govern execution senses.

Language: Английский

---

Memorability shapes perceived time (and vice versa)

Nature Human Behaviour, Journal Year: 2024, Volume and Issue: 8(7), P. 1296 - 1308

Published: April 22, 2024

Language: Английский

---

Benchmarking the speed–accuracy tradeoff in object recognition by humans and neural networks

Journal of Vision, Journal Year: 2025, Volume and Issue: 25(1), P. 4 - 4

Published: Jan. 3, 2025

Active object recognition, fundamental to tasks like reading and driving, relies on the ability make time-sensitive decisions. People exhibit a flexible tradeoff between speed accuracy, crucial human skill. However, current computational models struggle incorporate time. To address this gap, we present first dataset (with 148 observers) exploring speed–accuracy (SAT) in ImageNet recognition. Participants performed 16-way categorization task where their responses counted only if they occurred near time of fixed-delay beep. Each block trials allowed one reaction As expected, accuracy increases with We compare performance that dynamic neural networks adapt computation available inference Time is scarce resource for finding an appropriate analog challenging. Networks can repeat operations by using layers, recurrent cycles, or early exits. use repetition count as network's In our analysis, number exits correlates strongly floating-point operations, making them suitable analogs. Comparing humans SAT-fit error, category-wise correlation, SAT-curve steepness, find cascaded most promising modeling accuracy. Surprisingly, convolutional networks, typically favored recognition modeling, perform worst benchmark.

Language: Английский

---

Harnessing the power of artificial intelligence to transform hearing healthcare and research

Nature Machine Intelligence, Journal Year: 2021, Volume and Issue: 3(10), P. 840 - 849

Published: Oct. 18, 2021

Language: Английский

---

Predicting near-term, out-of-sample fish passage, guidance, and movement across diverse river environments by cognitively relating momentary behavioral decisions to multiscale memories of past hydrodynamic experiences

Frontiers in Ecology and Evolution, Journal Year: 2023, Volume and Issue: 11

Published: June 15, 2023

Predicting the behavior of individuals acting under their own motivation is a challenge shared across multiple scientific fields, from economic to ecological systems. In rivers, fish frequently change orientation even when stimuli are unchanged, which makes understanding and predicting movement in time-varying environments near built infrastructure particularly challenging. Cognition central movement, our lack costly terms time resources needed design manage water operations that able meet needs human society while preserving valuable living resources. An open question how best cognitively account for multi-modal, -attribute, -alternative, context-dependent decision-making infrastructure. Here, we leverage agent- individual-based modeling techniques encode cognitive approach mechanistic operates at scale river engineered managed. Our uses Eulerian-Lagrangian-agent method (ELAM) interpret quantitatively predict passage/entrainment different conditions. A goal methodology theory equations can provide an interpretable version animal complex requires minimal number parameters order facilitate application new data real-world engineering management projects. We first describe concepts, theory, mathematics applicable animals aquatic, terrestrial, avian, subterranean domains. Then, detail juvenile Pacific salmonids Bay-Delta California. reproduce observations salmon with one field season measurements, year 2009, using five simulated responses 3-D hydrodynamics. ELAM model calibrated 2009 data, later season, 2014, included novel guidance boom not present 2009. Central model’s performance notion attuned more than hydrodynamic signal timescale. find multi-timescale perception disentangle multiplex signals inform context-based behavioral choice fish. Simulated make decisions within rapidly changing environment without global information, knowledge direction downriver/upriver, or path integration. The key speed, spatial gradient acceleration, swim bladder pressure. selective tidal stream transport superset fish-hydrodynamic repertoire reproduces passage dam reservoir environments. From ecology perspective, emerge each tailored suit animal’s recent past experience (localized environmental context). paths same local stimuli. findings demonstrate does always require maximum possible spatiotemporal resolution representing although there concomitant tradeoffs resolving features scales. show decision-support tool successfully operate outside calibration conditions, necessary attribute tools informing future actions world will invariably look past.

Language: Английский

---

Boosting efficiency in a clinical literature surveillance system with LightGBM

PLOS Digital Health, Journal Year: 2024, Volume and Issue: 3(9), P. e0000299 - e0000299

Published: Sept. 23, 2024

Given the suboptimal performance of Boolean searching to identify methodologically sound and clinically relevant studies in large bibliographic databases, exploring machine learning (ML) efficiently classify is warranted. To boost efficiency a literature surveillance program, we used internationally recognized dataset articles tagged for methodological rigor applied an automated ML approach train test binary classification models predict probability clinical research being high methodologic quality. We trained over 12,000 on titles abstracts 97,805 indexed PubMed from 2012-2018 which were manually appraised by highly associates rated relevancy practicing clinicians. As unbalanced, with more that do not meet criteria rigor, unbalanced over- under-sampled datasets. Models maintained sensitivity at 99% maximized specificity selected tested retrospective set 30,424 2020 validated prospectively blinded study 5253 articles. The final algorithm, combining LightGBM (gradient boosting machine) model each dataset, achieved 57% validation 53% prospective study. number needed read find one met appraisal was 3.68 (95% CI 3.52 3.85) study, compared 4.63 4.50 4.77) when relying only searching. Gradient-boosting reduced work required quality 45%, improving subsequent dissemination clinicians other evidence users.

Language: Английский

---