A5003481800- Neural dynamics and brain function
- Neuroscience and Neuropharmacology Research
- Speech Recognition and Synthesis
- Music and Audio Processing
- Speech and Audio Processing
- Photoreceptor and optogenetics research
- Memory and Neural Mechanisms
- Visual perception and processing mechanisms
- Cognitive Science and Mapping
- Neural Networks and Applications
- Advanced Text Analysis Techniques
- Zebrafish Biomedical Research Applications
- Neuroscience and Neural Engineering
- Telemedicine and Telehealth Implementation
- Artificial Intelligence in Healthcare and Education
- Speech and dialogue systems
- Functional Brain Connectivity Studies
- Digital Mental Health Interventions
- Advanced Fluorescence Microscopy Techniques
- COVID-19 and healthcare impacts
- Advanced Memory and Neural Computing
- Blind Source Separation Techniques
- Machine Learning and Data Classification
- Animal Vocal Communication and Behavior
- Anomaly Detection Techniques and Applications
Google (United States)
2021-2025
Massachusetts Institute of Technology
2013-2021
Institute of Cognitive and Brain Sciences
2013-2021
Arden University
2021
McGovern Institute for Brain Research
2018-2019
Significance Understanding neurophysiological correlates of neurodevelopmental disorders is one the pressing challenges neuroscience. By analyzing a mouse model Rett syndrome (RTT), we show that cortical pyramidal neurons in methyl-CpG binding protein 2 (MeCP2) mutant mice have reduced excitatory as well inhibitory synaptic drive. Thus, neuronal response reliability and selectivity, features arise from excitatory/inhibitory processing circuits within cortex, are reduced. MeCP2 deletion...
Abstract Cognitive maps are mental representations of spatial and conceptual relationships in an environment, critical for flexible behavior. To form these abstract maps, the hippocampus has to learn separate or merge aliased observations appropriately different contexts a manner that enables generalization efficient planning. Here we propose specific higher-order graph structure, clone-structured cognitive (CSCG), which forms clones observation as representation addresses problems. CSCGs...
Intrinsic neuronal variability significantly limits information encoding in the primary visual cortex (V1). Certain stimuli can suppress this intertrial to increase reliability of responses. In particular, responses natural scenes, which have broadband spatiotemporal statistics, are more reliable than such as gratings. However, very little is known about stimulus statistics modulate coding and how occurs at neural ensemble level. Here, we sought elucidate role that spatial correlations...
Abstract The prefrontal cortex is vital for a range of cognitive processes, including working memory, attention, and decision-making. Notably, its absence impairs the performance tasks requiring maintenance information through delay period. In this paper, we formulate rodent task—which requires delay-period activity—as Markov decision process treat optimal task as an (active) inference problem. We simulate behavior Bayes mouse presented with 1 2 cues that instructs selection concurrent...
Intrinsic neuronal variability significantly limits information encoding in the primary visual cortex (V1). However, under certain conditions, neurons can respond reliably with highly precise responses to same stimuli from trial trial. This suggests that there exists intrinsic neural circuit mechanisms dynamically modulate intertrial of cortical neurons. Here, we sought elucidate role different inhibitory interneurons (INs) reliable coding mouse V1. To study interactions between...
Personalization of on-device speech recognition (ASR) has seen explosive growth in recent years, largely due to the increasing popularity personal assistant features on mobile devices and smart home speakers. In this work, we present Personal VAD 2.0, a personalized voice activity detector that detects target speaker, as part streaming ASR system. Although previous proof-of-concept studies have validated effectiveness VAD, there are still several critical challenges address before model can...
ObjectiveTo understand and highlight the differences in clinical, demographic, image quality characteristics between patient-taken (PAT) clinic-taken (CLIN) photographs of skin conditions.Patients MethodsThis retrospective study applied logistic regression to data from 2500 deidentified cases Stanford Health Care's eConsult system, November 2015 January 2021. Cases with undiagnosable or multiple conditions both patient clinician sources were excluded, leaving 628 PAT 1719 CLIN cases....
Abstract The uptake of glutamate by astrocytes actively shapes synaptic transmission, however its role in the development and plasticity neuronal circuits remains poorly understood. astrocytic transporter, GLT1 is predominant source clearance adult mouse cortex. Here, we examined structural functional visual cortex heterozygous (HET) mice using two‐photon microscopy, immunohistochemistry slice electrophysiology. We find that though eye‐specific thalamic axonal segregation intact, binocular...
In this paper, we introduce a streaming keyphrase detection system that can be easily customized to accurately detect any phrase composed of words from large vocabulary.The is implemented with an end-to-end trained automatic speech recognition (ASR) model and text-independent speaker verification model.To address the challenge detecting these keyphrases under various noisy conditions, separation added feature frontend model, adaptive noise cancellation (ANC) algorithm included exploit...
Accurate estimation of spike train from calcium (Ca <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2+</sup> ) fluorescence signals is challenging owing to significant fluctuations level. This paper proposes a non-model-based approach for inference using group delay (GD) analysis. It primarily exploits the property that change in Ca corresponding has notable onset location followed by decaying transient. The proposed algorithm, GDspike, compared...
Abstract Sequence learning is a vital cognitive function and has been observed in numerous brain areas. Discovering the algorithms underlying sequence major endeavour both neuroscience machine learning. In earlier work we showed that by constraining sparsity of emission matrix Hidden Markov Model (HMM) biologically-plausible manner are able to efficiently learn higher-order temporal dependencies recognize contexts noisy signals. The central basis our model, referred as Cloned HMM (CHMM),...
Abstract Cognitive maps are mental representations of spatial and conceptual relationships in an environment. These critical for flexible behavior as they permit us to navigate vicariously, but their underlying representation learning mechanisms still unknown. To form these abstract maps, hippocampus has learn separate or merge aliased observations appropriately different contexts a manner that enables generalization, efficient planning, handling uncertainty. Here we introduce specific...
In this paper, we propose a solution to allow speaker conditioned speech models, such as VoiceFilter-Lite, support an arbitrary number of enrolled users in single pass. This is achieved via attention mechanism on multiple embeddings compute attentive embedding, which then used side input the model. We implemented multi-user VoiceFilter-Lite and evaluated it for three tasks: (1) streaming automatic recognition (ASR) task; (2) text-independent verification (3) personalized keyphrase detection...
Abstract Hippocampus encodes cognitive maps that support episodic memories, navigation, and planning. Under-standing the commonality among those as well how are structured, learned from experience, used for inference planning is an interesting but unsolved problem. We propose higher-order graphs general principle present, a plausible model, cloned hidden Markov model (HMM) can learn these efficiently experienced sequences. In our experiments, we use HMM learning spatial abstract...
Abstract Human visual systems can parse a scene composed of novel objects and infer their surfaces occlusion relationships without relying on object-specific shapes or textures. Perceptual grouping bind together spatially disjoint entities to unite them as one object even when the is entirely novel, other perceptual properties like color texture that using object-based attention. Border-ownership assignment, assignment perceived boundaries specific surfaces, an intermediate representation in...
ABSTRACT Cortical neurons often respond to identical sensory stimuli with large variability. However, under certain conditions, the same can also highly reliably. The circuit mechanisms that contribute this modulation, and their influence on behavior remains unknown. Here we used novel double transgenic mice, dual-wavelength calcium imaging temporally selective optical perturbation identify an inhibitory neural in visual cortex modulate reliability of pyramidal naturalistic stimuli. Our...
VoiceFilter-Lite is a speaker-conditioned voice separation model that plays crucial role in improving speech recognition and speaker verification by suppressing overlapping from non-target speakers.However, one limitation of VoiceFilter-Lite, other models general, these are usually limited to single target speaker.This undesirable as most smart home devices now support multiple enrolled users.In order extend the benefits personalization users, we previously developed an attention-based...
Recently, Feature-wise Linear Modulation (FiLM) has been shown to outperform other approaches incorporate speaker embedding into speech separation and VoiceFilter models. We propose an improved method of incorporating such embeddings a Voice- Filter frontend for automatic recognition (ASR) text- independent verification (TI-SV). extend the widely- used Conformer architecture construct FiLM Block with additional feature processing before after layers. Apart from its application single-user...
Recently, there has been great progress in the ability of artificial intelligence (AI) algorithms to classify dermatological conditions from clinical photographs. However, little is known about robustness these real-world settings where several factors can lead a loss generalizability. Understanding and overcoming limitations will permit development generalizable AI that aid diagnosis skin across variety settings. In this retrospective study, we demonstrate differences condition...