Filters








5,297 Hits in 6.2 sec

Clustering units in neural networks: upstream vs downstream information [article]

Richard D. Lange, David S. Rolnick, Konrad P. Kording
2022 arXiv   pre-print
For each model, we quantify pairwise associations between hidden units in each layer using a variety of both upstream and downstream measures, then cluster them by maximizing their "modularity score" using  ...  It has been hypothesized that some form of "modular" structure in artificial neural networks should be useful for learning, compositionality, and generalization.  ...  These can be thought of as 2x2x2 product of methods, as shown in the color scheme in Figure 2 : the upstream vs downstream axis, the unnormalized vs normalized axis, and the covariance vs gradient (i.e  ... 
arXiv:2203.11815v1 fatcat:hakgvrcidzakdov45ddktl3wwm

Built to Last: Functional and structural mechanisms in the moth olfactory network mitigate effects of neural injury [article]

Charles B Delahunt, Pedro D Maia, J. Nathan Kutz
2020 arXiv   pre-print
Axonal swellings effectively compromise spike train propagation along the axon, reducing the effective neural firing rate delivered to downstream neurons.  ...  We simulate injuries on a detailed computational model of the moth olfactory network calibrated to in vivo data.  ...  Fig. 9 9 Injury mitigation hypotheses: In a cascaded network, various architectures can mitigate the effects of injury to upstream neurons by protecting or restoring functionality of downstream units.  ... 
arXiv:1808.01279v3 fatcat:kbu42wa3uzbhbppxq6zry7w7aq

Built to Last: Functional and Structural Mechanisms in the Moth Olfactory Network Mitigate Effects of Neural Injury

Charles B. Delahunt, Pedro D. Maia, J. Nathan Kutz
2021 Brain Sciences  
Axonal swellings effectively compromise spike train propagation along the axon, reducing the effective neural firing rate delivered to downstream neurons.  ...  In this work, we explore whether and how certain structural and functional neuronal network motifs act as injury mitigation mechanisms.  ...  Injury mitigation hypotheses: In a cascaded network, various architectures can mitigate the effects of injury to upstream neurons by protecting or restoring functionality of downstream units.  ... 
doi:10.3390/brainsci11040462 pmid:33916469 pmcid:PMC8067361 fatcat:3xug45lb5ngbjgmbi3twafwrie

Geometric Classification of Brain Network Dynamics via Conic Derivative Discriminants [article]

Matthew F. Singh, Todd Braver, ShiNung Ching
2017 bioRxiv   pre-print
By doing so, dynamically-coded information may be revealed in terms of geometric patterns in the phase space of the derivative signal.  ...  Over the past decade, pattern decoding techniques have granted neuroscientists improved anatomical specificity in mapping neural representations associated with function and cognition.  ...  we performed unsupervised clustering of downstream neural voltage.  ... 
doi:10.1101/201905 fatcat:wahouniunbaj7aa63wqxg5ofau

Geometric classification of brain network dynamics via conic derivative discriminants

Matthew F. Singh, Todd S. Braver, ShiNung Ching
2018 Journal of Neuroscience Methods  
Over the past decade, pattern decoding techniques have granted neuroscientists improved anatomical specificity in mapping neural representations associated with function and cognition.  ...  we performed unsupervised clustering of downstream neural voltage.  ...  Thus, by altering the activity of a mediating cell, the upstream component's activation dictates possible interactions in the downstream network.  ... 
doi:10.1016/j.jneumeth.2018.06.019 pmid:29966600 pmcid:PMC6417100 fatcat:tvbw6xw6mjhtbkqktzc6owm4ja

MethylSPWNet and MethylCapsNet: Biologically Motivated Organization of DNAm Neural Networks, Inspired by Capsule Networks

Joshua J. Levy, Youdinghuan Chen, Nasim Azizgolshani, Curtis L. Petersen, Alexander J. Titus, Erika L. Moen, Louis J. Vaickus, Lucas A. Salas, Brock C. Christensen
2021 npj Systems Biology and Applications  
AbstractDNA methylation (DNAm) alterations have been heavily implicated in carcinogenesis and the pathophysiology of diseases through upstream regulation of gene expression.  ...  We demonstrate these models' utility on 3,897 individuals in the classification of central nervous system (CNS) tumors.  ...  CLP and JJL are supported through the Burroughs Wellcome Fund Big Data in the Life Sciences at Dartmouth.  ... 
doi:10.1038/s41540-021-00193-7 pmid:34417465 pmcid:PMC8379254 fatcat:gehkq2j53vbo3njmuj2s5mib2q

A mechanistic multi-area recurrent network model of decision-making

Michael Kleinman, Chandramouli Chandrasekaran, Jonathan C. Kao
2021 Neural Information Processing Systems  
Recurrent neural networks (RNNs) trained on neuroscience-based tasks have been widely used as models for cortical areas performing analogous tasks.  ...  In particular, we show that incorporating multiple areas and Dale's Law is critical for biasing the networks to learn biologically plausible solutions.  ...  Chandrasekaran when he was a postdoc in the Shenoy Lab. MK was supported by the National Sciences and Engineering Research Council (NSERC).  ... 
dblp:conf/nips/KleinmanCK21 fatcat:whkhfpopwbfvfbovgjp3zuqnfy

Prediction of Data Traffic in Telecom Networks based on Deep Neural Networks

Quang Hung Do, Thi Thanh Hang Doan, Thi Van Anh Nguyen, Nguyen Tung Duong, Vu Van Linh
2020 Journal of Computer Science  
This study employs deep neural networks including Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) techniques to one-hour-ahead forecast the volume of expected traffic and compares this approach  ...  The data traffic was collected from EVN Telecom mobile communication network. The results showed that shows a good estimation when dealing with volatility clustering in the data series.  ...  LSTM has a unit of computation, a memory cell that replaces traditional artificial neurons in the hidden layer of the neural network.  ... 
doi:10.3844/jcssp.2020.1268.1277 fatcat:inolodbyzjdunmkhr3oyrqumui

Machine learning of stochastic gene network phenotypes [article]

Kyemyung Park, Thorsten Prustel, Yong Lu, John S Tsang
2019 bioRxiv   pre-print
We applied our approach to investigate stochastic gene expression propagation in biological networks, which is a contemporary challenge in the quantitative modeling of single-cell heterogeneity.  ...  Even in the simplest networks existing analytical schemes generated significantly less accurate predictions than our approach, which revealed interesting insights when applied to more complex circuits,  ...  (F) Illustrating the role of ; (; &( , ; (: , and ; &: ), the level of upstream input needed to attain half maximal activation of the downstream gene, plays in tuning the propagation of variability/information  ... 
doi:10.1101/825943 fatcat:e2jp6zqfnbhjlbvnbnav5yq6yy

A ROLLING-TRAINED FUZZY NEURAL NETWORK APPROACH FOR FREEWAY INCIDENT DETECTION

Lawrence W. Lan, Yeh-Chieh Huang
2006 Transportmetrica  
The core logic of this approach is to establish a fuzzy neural network and to update the network parameters in response to the prevailing traffic conditions through a rolling-trained procedure.  ...  This paper develops a rolling-trained fuzzy neural network (RTFNN) approach for freeway incident detection.  ...  The gate network classifies the inputs into several clusters using a fuzzy approach and the expert network specifies the input-output relationship as in a conventional NN approach.  ... 
doi:10.1080/18128600608685653 fatcat:scelilnyebfrxlslboz2hsnbfm

What Do Neural Networks Learn When Trained With Random Labels? [article]

Hartmut Maennel and Ibrahim Alabdulmohsin and Ilya Tolstikhin and Robert J. N. Baldock and Olivier Bousquet and Sylvain Gelly and Daniel Keysers
2020 arXiv   pre-print
We study deep neural networks (DNNs) trained on natural image data with entirely random labels.  ...  We show how this alignment produces a positive transfer: networks pre-trained with random labels train faster downstream compared to training from scratch even after accounting for simple effects, such  ...  Nevertheless, we hope that a better understanding of deep neural networks will lead to improvements in the future along the direction of building interpretable and explainable AI, which are critical ingredients  ... 
arXiv:2006.10455v2 fatcat:hk2y5zpubnbkzagbourle7xtqa

Foxm1 controls a pro-stemness microRNA network in neural stem cells

Zein Mersini Besharat, Luana Abballe, Francesco Cicconardi, Arjun Bhutkar, Luigi Grassi, Loredana Le Pera, Marta Moretti, Mauro Chinappi, Daniel D'Andrea, Angela Mastronuzzi, Alessandra Ianari, Alessandra Vacca (+5 others)
2018 Scientific Reports  
Genes with higher transcript levels in NSCs (vs. Diff-NSCs) included Foxm1, which proved to be directly regulated by Gli and Nanog.  ...  Foxm1 in turn regulated several microRNAs that were overexpressed in NSCs: miR-130b, miR-301a, and members of the miR-15~16 and miR-17~92 clusters and whose knockdown significantly impaired the neurosphere  ...  As shown in Fig. 6A , four putative binding sites for Nanog were found −3790 to −3277 bp upstream from the Foxm1 TSS. (For details, see Supplementary Information).  ... 
doi:10.1038/s41598-018-21876-y pmid:29476172 pmcid:PMC5824884 fatcat:i56qakqohfdknas6lutjqngfru

Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Ventral Stream

U. Guclu, M. A. J. van Gerven
2015 Journal of Neuroscience  
Using deep convolutional neural networks, we can now quantitatively demonstrate that there is indeed an explicit gradient for feature complexity in the ventral pathway of the human brain.  ...  Finally, it is shown that deep convolutional neural networks allow decoding of representations in the human brain at a previously unattainable degree of accuracy, providing a more sensitive window into  ...  This is consistent with neurophysiological findings in primates that some downstream neurons are tuned to relatively simple features and some upstream neurons are tuned to relatively complex features  ... 
doi:10.1523/jneurosci.5023-14.2015 pmid:26157000 fatcat:earigmu2irhkbiprv7pnsar7ze

Greenhouse Gas Emission Prediction on Road Network using Deep Sequence Learning [article]

Lama Alfaseeh, Ran Tu, Bilal Farooq, Marianne Hatzopoulou
2020 arXiv   pre-print
In particular, various specifications of the long-short term memory (LSTM) networks with exogenous variables are examined and compared with clustering and the autoregressive integrated moving average (  ...  The downtown Toronto road network is used as the case study and highly detailed data are synthesized using a calibrated traffic microsimulation and MOVES.  ...  In addition, they found that their proposed approach outperformed the ELM and the back propagation neural network in terms of the RMSE and MAPE.  ... 
arXiv:2004.08286v1 fatcat:ztoirhtk2nc2dnj5ll5qad626e

Network Dynamics Underlying the Formation of Sparse, Informative Representations in the Hippocampus

M. P. Karlsson, L. M. Frank
2008 Journal of Neuroscience  
This process is not evident in CA3, indicating that a region-specific and long timescale process operates in CA1 to create a sparse, spatially informative population of neurons.  ...  Here we show that different dynamics govern the evolution of this sparsity in CA1 and upstream area CA3. CA1, but not CA3, produces twice as many spikes in novel compared with familiar environments.  ...  After neural data were collected, individual units were identified by clustering spikes using only peak amplitude and spike width as variables.  ... 
doi:10.1523/jneurosci.4261-08.2008 pmid:19109508 pmcid:PMC2632980 fatcat:a5pdnnyds5d6fhttkz6laujksu
« Previous Showing results 1 — 15 out of 5,297 results