Noise correlations for faster and more robust learning

Mapping Intimacies ◽

10.1101/2020.10.15.341768 ◽

2020 ◽

Author(s):

Matthew R. Nassar ◽

Apoorva Bhandari

Keyword(s):

Hebbian Learning ◽

Signal To Noise Ratio ◽

Signal To Noise ◽

Learning Problem ◽

Robust Learning ◽

Irrelevant Feature ◽

Population Codes ◽

Noise Ratio ◽

Stored Information ◽

The Brain

AbstractDistributed population codes are ubiquitous in the brain and pose a challenge to downstream neurons that must learn an appropriate readout. Here we explore the possibility that this learning problem is simplified through inductive biases implemented by stimulus-independent noise correlations that constrain learning to task-relevant dimensions. We test this idea in a set of neural networks that learn to perform a perceptual discrimination task. Correlations among similarly tuned units were manipulated independently of overall population signal-to-noise ratio in order to test how the format of stored information affects learning. Higher noise correlations among similarly tuned units led to faster and more robust learning, favoring homogenous weights assigned to neurons within a functionally similar pool, and could emerge through Hebbian learning. When multiple discriminations were learned simultaneously, noise correlations across relevant feature dimensions sped learning whereas those across irrelevant feature dimensions slowed it. Our results complement existing theory on noise correlations by demonstrating that when such correlations are produced without degradation of signal-to-noise ratio, they can improve readout learning by constraining it to appropriate dimensions.

Download Full-text

Automated segmentation of thick confocal microscopy 3D images for the measurement of white matter volumes in zebrafish brains

Mathematical Morphology - Theory and Applications ◽

10.1515/mathm-2020-0100 ◽

2020 ◽

Vol 4 (1) ◽

pp. 31-45 ◽

Cited By ~ 1

Author(s):

Sylvain Lempereur ◽

Arnim Jenett ◽

Elodie Machado ◽

Ignacio Arganda-Carreras ◽

Matthieu Simion ◽

...

Keyword(s):

White Matter ◽

Signal To Noise Ratio ◽

Light Attenuation ◽

Automated Image Analysis ◽

High Signal ◽

Signal To Noise ◽

Tissue Clearing ◽

Brain White Matter ◽

Noise Ratio ◽

The Brain

AbstractTissue clearing methods have boosted the microscopic observations of thick samples such as whole-mount mouse or zebrafish. Even with the best tissue clearing methods, specimens are not completely transparent and light attenuation increases with depth, reducing signal output and signal-to-noise ratio. In addition, since tissue clearing and microscopic acquisition techniques have become faster, automated image analysis is now an issue. In this context, mounting specimens at large scale often leads to imperfectly aligned or oriented samples, which makes relying on predefined, sample-independent parameters to correct signal attenuation impossible.Here, we propose a sample-dependent method for contrast correction. It relies on segmenting the sample, and estimating sample depth isosurfaces that serve as reference for the correction. We segment the brain white matter of zebrafish larvae. We show that this correction allows a better stitching of opposite sides of each larva, in order to image the entire larva with a high signal-to-noise ratio throughout. We also show that our proposed contrast correction method makes it possible to better recognize the deep structures of the brain by comparing manual vs. automated segmentations. This is expected to improve image observations and analyses in high-content methods where signal loss in the samples is significant.

Download Full-text

The dynamics of explore-exploit decisions reveal a signal-to-noise mechanism for random exploration

10.31234/osf.io/uepr7 ◽

2020 ◽

Author(s):

Samuel Franklin Feng ◽

Siyu Wang ◽

Sylvia Zarnescu ◽

Robert C Wilson

Keyword(s):

Drift Rate ◽

Signal To Noise Ratio ◽

Critical Role ◽

Behavioral Variability ◽

Drift Diffusion Model ◽

Signal To Noise ◽

Noise Ratio ◽

Reward Information ◽

Model Variability ◽

The Brain

Growing evidence suggests that behavioral variability plays a critical role in how humans manage the trade-off between exploration and exploitation. In these decisions a little variability can help us to overcome the desire to exploit known rewards by encouraging us to randomly explore something else. Here we investigate how such `random exploration' could be controlled using a drift-diffusion model of the explore-exploit choice. In this model, variability is controlled by either the signal-to-noise ratio with which reward is encoded (the `drift rate'), or the amount of information required before a decision is made (the `threshold'). By fitting this model to behavior, we find that while, statistically, both drift and threshold change when people randomly explore, numerically, the change in drift rate has by far the largest effect. This suggests that random exploration is primarily driven by changes in the signal-to-noise ratio with which reward information is represented in the brain.

Download Full-text

Signal-to-noise ratio and dose to the lens of the eye for computed tomography examination of the brain using an automatic tube current modulation system

Emergency Radiology ◽

10.1007/s10140-016-1470-6 ◽

2016 ◽

Vol 24 (3) ◽

pp. 233-239 ◽

Cited By ~ 3

Author(s):

Supawitoo Sookpeng ◽

Chitsanupong Butdee

Keyword(s):

Computed Tomography ◽

Signal To Noise Ratio ◽

Tube Current Modulation ◽

Signal To Noise ◽

Current Modulation ◽

Automatic Tube Current Modulation ◽

Tube Current ◽

Noise Ratio ◽

The Brain

Download Full-text

The dynamics of explore–exploit decisions reveal a signal-to-noise mechanism for random exploration

Scientific Reports ◽

10.1038/s41598-021-82530-8 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Samuel F. Feng ◽

Siyu Wang ◽

Sylvia Zarnescu ◽

Robert C. Wilson

Keyword(s):

Drift Rate ◽

Signal To Noise Ratio ◽

Critical Role ◽

Behavioral Variability ◽

Drift Diffusion Model ◽

Signal To Noise ◽

Noise Ratio ◽

Reward Information ◽

Model Variability ◽

The Brain

AbstractGrowing evidence suggests that behavioral variability plays a critical role in how humans manage the tradeoff between exploration and exploitation. In these decisions a little variability can help us to overcome the desire to exploit known rewards by encouraging us to randomly explore something else. Here we investigate how such ‘random exploration’ could be controlled using a drift-diffusion model of the explore–exploit choice. In this model, variability is controlled by either the signal-to-noise ratio with which reward is encoded (the ‘drift rate’), or the amount of information required before a decision is made (the ‘threshold’). By fitting this model to behavior, we find that while, statistically, both drift and threshold change when people randomly explore, numerically, the change in drift rate has by far the largest effect. This suggests that random exploration is primarily driven by changes in the signal-to-noise ratio with which reward information is represented in the brain.

Download Full-text

Determination of the Best Emulsion for the Microscopy of Crystals

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100096886 ◽

1981 ◽

Vol 39 ◽

pp. 32-33

Author(s):

David A. Grano ◽

Kenneth H. Downing

Keyword(s):

Spatial Frequency ◽

Information Transfer ◽

Signal To Noise Ratio ◽

Experimental Situation ◽

Photographic Emulsion ◽

Modulation Transfer ◽

Signal To Noise ◽

Frequency Space ◽

Noise Ratio

The retrieval of high-resolution information from images of biological crystals depends, in part, on the use of the correct photographic emulsion. We have been investigating the information transfer properties of twelve emulsions with a view toward 1) characterizing the emulsions by a few, measurable quantities, and 2) identifying the “best” emulsion of those we have studied for use in any given experimental situation. Because our interests lie in the examination of crystalline specimens, we've chosen to evaluate an emulsion's signal-to-noise ratio (SNR) as a function of spatial frequency and use this as our critereon for determining the best emulsion.The signal-to-noise ratio in frequency space depends on several factors. First, the signal depends on the speed of the emulsion and its modulation transfer function (MTF). By procedures outlined in, MTF's have been found for all the emulsions tested and can be fit by an analytic expression 1/(1+(S/S0)2). Figure 1 shows the experimental data and fitted curve for an emulsion with a better than average MTF. A single parameter, the spatial frequency at which the transfer falls to 50% (S0), characterizes this curve.

Download Full-text

Experiments in Bright-Field Imaging with Hollow-Cone Illumination

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100097703 ◽

1981 ◽

Vol 39 ◽

pp. 226-227

Author(s):

W. Kunath ◽

K. Weiss ◽

E. Zeitler

Keyword(s):

Signal To Noise Ratio ◽

Carbon Support ◽

Electronic System ◽

Optic Axis ◽

Hollow Cone ◽

Signal To Noise ◽

Bright Field ◽

Noise Ratio ◽

Single Field ◽

Field Imaging

Bright-field images taken with axial illumination show spurious high contrast patterns which obscure details smaller than 15 ° Hollow-cone illumination (HCI), however, reduces this disturbing granulation by statistical superposition and thus improves the signal-to-noise ratio. In this presentation we report on experiments aimed at selecting the proper amount of tilt and defocus for improvement of the signal-to-noise ratio by means of direct observation of the electron images on a TV monitor.Hollow-cone illumination is implemented in our microscope (single field condenser objective, Cs = .5 mm) by an electronic system which rotates the tilted beam about the optic axis. At low rates of revolution (one turn per second or so) a circular motion of the usual granulation in the image of a carbon support film can be observed on the TV monitor. The size of the granular structures and the radius of their orbits depend on both the conical tilt and defocus.

Download Full-text

Information capacity and bandwidth limitations in the SEM

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100152392 ◽

1989 ◽

Vol 47 ◽

pp. 84-85

Author(s):

D. C. Joy ◽

R. D. Bunn

Keyword(s):

Signal To Noise Ratio ◽

Critical Dimension ◽

Information Capacity ◽

Acquisition Time ◽

Signal To Noise ◽

Sem Image ◽

Probe Size ◽

Minimum Number ◽

Noise Ratio ◽

Signal Channel

The information available from an SEM image is limited both by the inherent signal to noise ratio that characterizes the image and as a result of the transformations that it may undergo as it is passed through the amplifying circuits of the instrument. In applications such as Critical Dimension Metrology it is necessary to be able to quantify these limitations in order to be able to assess the likely precision of any measurement made with the microscope.The information capacity of an SEM signal, defined as the minimum number of bits needed to encode the output signal, depends on the signal to noise ratio of the image - which in turn depends on the probe size and source brightness and acquisition time per pixel - and on the efficiency of the specimen in producing the signal that is being observed. A detailed analysis of the secondary electron case shows that the information capacity C (bits/pixel) of the SEM signal channel could be written as :

Download Full-text

Speech Reception in the Presence of Classroom Noise

Language Speech and Hearing Services in Schools ◽

10.1044/0161-1461.1004.221 ◽

1979 ◽

Vol 10 (4) ◽

pp. 221-230 ◽

Cited By ~ 2

Author(s):

Veronica Smyth

Keyword(s):

Signal To Noise Ratio ◽

Learning Processes ◽

Speech Discrimination ◽

Signal To Noise ◽

Speech Reception ◽

Monosyllabic Words ◽

Noise Ratio

Three hundred children from five to 12 years of age were required to discriminate simple, familiar, monosyllabic words under two conditions: 1) quiet, and 2) in the presence of background classroom noise. Of the sample, 45.3% made errors in speech discrimination in the presence of background classroom noise. The effect was most marked in children younger than seven years six months. The results are discussed considering the signal-to-noise ratio and the possible effects of unwanted classroom noise on learning processes.

Download Full-text

Quantifying the Effects of Background Speech Babble on Preschool Children's Novel Word Learning in a Multisession Paradigm: A Preliminary Study

Journal of Speech Language and Hearing Research ◽

10.1044/2019_jslhr-h-19-0083 ◽

2020 ◽

Vol 63 (1) ◽

pp. 345-356

Author(s):

Meital Avivi-Reich ◽

Megan Y. Roberts ◽

Tina M. Grieco-Calub

Keyword(s):

Word Learning ◽

Signal To Noise Ratio ◽

Preschool Age ◽

Exposure Condition ◽

Signal To Noise ◽

Verbal Recall ◽

Referent Selection ◽

Main Effect ◽

Noise Ratio ◽

Novel Word Learning

Purpose This study tested the effects of background speech babble on novel word learning in preschool children with a multisession paradigm. Method Eight 3-year-old children were exposed to a total of 8 novel word–object pairs across 2 story books presented digitally. Each story contained 4 novel consonant–vowel–consonant nonwords. Children were exposed to both stories, one in quiet and one in the presence of 4-talker babble presented at 0-dB signal-to-noise ratio. After each story, children's learning was tested with a referent selection task and a verbal recall (naming) task. Children were exposed to and tested on the novel word–object pairs on 5 separate days within a 2-week span. Results A significant main effect of session was found for both referent selection and verbal recall. There was also a significant main effect of exposure condition on referent selection performance, with more referents correctly selected for word–object pairs that were presented in quiet compared to pairs presented in speech babble. Finally, children's verbal recall of novel words was statistically better than baseline performance (i.e., 0%) on Sessions 3–5 for words exposed in quiet, but only on Session 5 for words exposed in speech babble. Conclusions These findings suggest that background speech babble at 0-dB signal-to-noise ratio disrupts novel word learning in preschool-age children. As a result, children may need more time and more exposures of a novel word before they can recognize or verbally recall it.

Download Full-text