scholarly journals Two sides of the same coin: Beneficial and detrimental consequences of range adaptation in human reinforcement learning

2021 ◽  
Vol 7 (14) ◽  
pp. eabe0340
Author(s):  
Sophie Bavard ◽  
Aldo Rustichini ◽  
Stefano Palminteri
Keyword(s):  
Reinforcement Learning ◽  
Task Difficulty ◽  
Signal To Noise Ratio ◽  
Signal To Noise ◽  
Economic Values ◽  
Adaptive Coding ◽  
Large Sample ◽  
Locally Adaptive ◽  
And Performance ◽  
Two Sides

Evidence suggests that economic values are rescaled as a function of the range of the available options. Although locally adaptive, range adaptation has been shown to lead to suboptimal choices, particularly notable in reinforcement learning (RL) situations when options are extrapolated from their original context to a new one. Range adaptation can be seen as the result of an adaptive coding process aiming at increasing the signal-to-noise ratio. However, this hypothesis leads to a counterintuitive prediction: Decreasing task difficulty should increase range adaptation and, consequently, extrapolation errors. Here, we tested the paradoxical relation between range adaptation and performance in a large sample of participants performing variants of an RL task, where we manipulated task difficulty. Results confirmed that range adaptation induces systematic extrapolation errors and is stronger when decreasing task difficulty. Last, we propose a range-adapting model and show that it is able to parsimoniously capture all the behavioral results.

scholarly journals The construction and deconstruction of sub-optimal preferences through range-adapting reinforcement learning

2020 ◽  
Author(s):  
Sophie Bavard ◽  
Aldo Rustichini ◽  
Stefano Palminteri
Keyword(s):  
Reinforcement Learning ◽  
Task Difficulty ◽  
Signal To Noise Ratio ◽  
Signal To Noise ◽  
Economic Values ◽  
Adaptive Coding ◽  
Large Sample ◽  
Locally Adaptive ◽  
Noise Ratio ◽  
And Performance

Converging evidence suggests that economic values are rescaled as a function of the range of the available options. Critically, although locally adaptive, range adaptation has been shown to lead to suboptimal choices. This is particularly striking in reinforcement learning (RL) situations when options are extrapolated from their original context. Range adaptation can be seen as the result of an adaptive coding process aiming at increasing the signal-to-noise ratio. However, this hypothesis leads to a counterintuitive prediction: decreasing outcome uncertainty should increase range adaptation and, consequently, extrapolation errors. Here, we tested the paradoxical relation between range adaptation and performance in a large sample of subjects performing variants of a RL task, where we manipulated task difficulty. Results confirmed that range adaptation induces systematic extrapolation errors and is stronger when decreasing outcome uncertainty. Finally, we propose a range-adapting model and show that it is able to parsimoniously capture all the observed results.

scholarly journals The construction and deconstruction of sub-optimal preferences through range-adapting reinforcement learning

2020 ◽  
Author(s):  
Sophie Bavard ◽  
Aldo Rustichini ◽  
Stefano Palminteri
Keyword(s):  
Reinforcement Learning ◽  
Task Difficulty ◽  
Signal To Noise Ratio ◽  
Signal To Noise ◽  
Economic Values ◽  
Adaptive Coding ◽  
Large Sample ◽  
Locally Adaptive ◽  
Noise Ratio ◽  
And Performance

AbstractConverging evidence suggests that economic values are rescaled as a function of the range of the available options. Critically, although locally adaptive, range adaptation has been shown to lead to suboptimal choices. This is particularly striking in reinforcement learning (RL) situations when options are extrapolated from their original context. Range adaptation can be seen as the result of an adaptive coding process aiming at increasing the signal-to-noise ratio. However, this hypothesis leads to a counter-intuitive prediction: decreasing outcome uncertainty should increase range adaptation and, consequently, extrapolation errors. Here, we tested the paradoxical relation between range adaptation and performance in a large sample of subjects performing variants of a RL task, where we manipulated task difficulty. Results confirmed that range adaptation induces systematic extrapolation errors and is stronger when decreasing outcome uncertainty. Finally, we propose a range-adapting model and show that it is able to parsimoniously capture all the observed results.

Peak Signal-to-Noise Ratio Evaluation of Server Display Monitors and Client Display Monitors in a Digital Subtraction Angiography Devices

2020 ◽  
Vol 3 (1) ◽  
pp. 33-41
Author(s):  
Hwunjae Lee ◽  
◽  
Junhaeng Lee ◽  
Keyword(s):  
Signal To Noise Ratio ◽  
Information Loss ◽  
The Other ◽  
Digital Subtraction ◽  
Signal Distortion ◽  
Signal To Noise ◽  
Image Storage ◽  
And Performance ◽  
Pacs Server ◽  
Display Monitor

This study evaluated PSNR of server display monitor and client display monitor of DSA system. The signal is acquired and imaged during the surgery and stored in the PACS server. After that, distortion of the original signal is an important problem in the process of observation on the client monitor. There are many problems such as noise generated during compression and image storage/transmission in PACS, information loss during image storage and transmission, and deterioration in image quality when outputting medical images from a monitor. The equipment used for the experiment in this study was P's DSA. We used two types of monitors in our experiment, one is P’s company resolution 1280×1024 pixel monitor, and the other is W’s company resolution 1536×2048 pixel monitor. The PACS Program used MARO-view, and for the experiment, a PSNR measurement program using Visual C++ was implemented and used for the experiment. As a result of the experiment, the PSNR value of the kidney angiography image was 26.958dB, the PSNR value of the lung angiography image was 28.9174 dB, the PSNR value of the heart angiography image was 22.8315dB, and the PSNR value of the neck angiography image was 37.0319 dB, and the knee blood vessels image showed a PSNR value of 43.2052 dB, respectively. In conclusion, it can be seen that there is almost no signal distortion in the process of acquiring, storing, and transmitting images in PACS. However, it suggests that the image signal may be distorted depending on the resolution and performance of each monitor. Therefore, it will be necessary to evaluate the performance of the monitor and to maintain the performance.

In situ studies of metal–semiconductor interactions with synchrotron radiation

1998 ◽  
Vol 5 (3) ◽  
pp. 1050-1051
Author(s):  
D. E. Sayers ◽  
P. T. Goeller ◽  
B. I. Boyanov ◽  
R. J. Nemanich
Keyword(s):  
Synchrotron Radiation ◽  
Silicon Germanium ◽  
Signal To Noise Ratio ◽  
Thick Films ◽  
Strained Silicon ◽  
Reaction Products ◽  
Signal To Noise ◽  
In Situ Studies ◽  
And Performance

The capabilities and performance of a UHV system for in situ studies of metal–semiconductor interactions are described. The UHV system consists of interconnected deposition and analysis chambers, each of which is capable of maintaining a base pressure of approximately 1 × 10−10 torr. The deposited materials and their reaction products can be studied in situ with RHEED, XAFS, AES, XPS, UPS and ARUPS. Results from a study of the reaction of 0.7- and 1.7-monolayer-thick films of cobalt with strained silicon–germanium alloys are presented. The signal-to-noise ratio obtained in these experiments indicates that the apparatus is capable of supporting in situ EXAFS studies of ∼0.1-monolayer-thick films.

Enhancement of Gradient for Touching and Overlapping Objects

2020 ◽  
Vol 17 (2) ◽  
pp. 1173-1176
Author(s):  
Wan Mahani Abdullah ◽  
Shahrul Nizam Yaakob ◽  
Ainul Maulid Ahmad ◽  
Mu’azah Md Aziz ◽  
Mohamad Izril Ishak ◽  
...  
Keyword(s):  
Mean Squared Error ◽  
Signal To Noise Ratio ◽  
Signal To Noise ◽  
Gradient Image ◽  
Squared Error ◽  
Complex Background ◽  
Performance Matrix ◽  
The Mean ◽  
And Performance ◽  
Gradient Enhancement

Normally, gradient image is sometime poor especially for touching and overlapping objects. The proposed method manipulates the use of HSV channels in order to enhance the gradient image obtained for objects with complex background. Only S and V channels are used in this technique. The proposed gradient enhancement approaches were tested on HSV channel images. The Standard Sobel and Laplace of Gaussian (LoG) technique was used to compare the results and performance. The Mean Squared Error (MSE) and Peak signal-to-noise ratio (PSNR) values were calculated as a performance matrix for the implementation. The results show the improvement of the gradient image if compared to existence approach.

scholarly journals Acoustic-Perceptual Factors Both Maintain and Account for the Rarity of the Czech Trill-Fricative

2022 ◽  
Vol 6 ◽  
Author(s):  
Phil J. Howson ◽  
Philip J. Monahan
Keyword(s):  
Multidimensional Scaling ◽  
Task Difficulty ◽  
Discrimination Task ◽  
Signal To Noise Ratio ◽  
Scaling Analysis ◽  
Perceptual Cues ◽  
Signal To Noise ◽  
Multidimensional Scaling Analysis ◽  
Perceptual Characteristics ◽  
Perceptual Factors

Czech has a sibilant inventory that contrasts at three places of articulation: Alveolar, a pre-post-alveolar, and palato-alveolar. The specific aim of this study is to examine the perception of the typologically rare Czech sibilant inventory and to determine whether acoustic-perceptual characteristics play a role in the maintenance of the Czech trill-fricative. These results are compared to a more common three-way sibilant inventory, Polish. Native Czech listeners performed an auditory AX discrimination task in two blocks: A Czech block and a Polish block. Stimuli were embedded in varying levels of noise to increase task difficulty. Signal-to-noise ratio differences affected the perception of the Czech sibilants more than Polish sibilants. Moreover, a multidimensional scaling analysis revealed less perceptual dispersion for the Czech inventory than the Polish inventory. These results suggest that there is greater difficulty maintaining the Czech inventory considering the signal-to-noise comparisons and that this a factor that contributes to its rarity; however, similarities in perceptual dispersion indicate that maintenance across several acoustic-perceptual cues is possible, and Czech shows few signs of losing this typologically rare contrast.

Sign in / Sign up

Export Citation Format

Share Document