5 results
Clarifying status of DNNs as models of human vision
- Jeffrey S. Bowers, Gaurav Malhotra, Marin Dujmović, Milton L. Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E. Hummel, Rachel F. Heaton, Benjamin D. Evans, Jeffrey Mitchell, Ryan Blything
-
- Journal:
- Behavioral and Brain Sciences / Volume 46 / 2023
- Published online by Cambridge University Press:
- 06 December 2023, e415
-
- Article
- Export citation
-
On several key issues we agree with the commentators. Perhaps most importantly, everyone seems to agree that psychology has an important role to play in building better models of human vision, and (most) everyone agrees (including us) that deep neural networks (DNNs) will play an important role in modelling human vision going forward. But there are also disagreements about what models are for, how DNN–human correspondences should be evaluated, the value of alternative modelling approaches, and impact of marketing hype in the literature. In our view, these latter issues are contributing to many unjustified claims regarding DNN–human correspondences in vision and other domains of cognition. We explore all these issues in this response.
Deep problems with neural network models of human vision
- Jeffrey S. Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E. Hummel, Rachel F. Heaton, Benjamin D. Evans, Jeffrey Mitchell, Ryan Blything
-
- Journal:
- Behavioral and Brain Sciences / Volume 46 / 2023
- Published online by Cambridge University Press:
- 01 December 2022, e385
-
- Article
- Export citation
-
Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models of biological vision. This conclusion is largely based on three sets of findings: (1) DNNs are more accurate than any other model in classifying images taken from various datasets, (2) DNNs do the best job in predicting the pattern of human errors in classifying objects taken from various behavioral datasets, and (3) DNNs do the best job in predicting brain signals in response to images taken from various brain datasets (e.g., single cell responses or fMRI data). However, these behavioral and brain datasets do not test hypotheses regarding what features are contributing to good predictions and we show that the predictions may be mediated by DNNs that share little overlap with biological vision. More problematically, we show that DNNs account for almost no results from psychological research. This contradicts the common claim that DNNs are good, let alone the best, models of human object recognition. We argue that theorists interested in developing biologically plausible models of human vision need to direct their attention to explaining psychological findings. More generally, theorists need to build models that explain the results of experiments that manipulate independent variables designed to test hypotheses rather than compete on making the best predictions. We conclude by briefly summarizing various promising modeling approaches that focus on psychological data.
IR hot carrier based photodetection in titanium nitride oxide thin film-Si junctions
- Nicholas A. Güsken, Alberto Lauri, Yi Li, Andrea Jacassi, Takayuki Matsui, Brock Doiron, Ryan Bower, Anna Regoutz, Andrei Mihai, Peter K. Petrov, Rupert F. Oulton, Lesley F. Cohen, Stefan A. Maier
-
- Journal:
- MRS Advances / Volume 5 / Issue 35-36 / 2020
- Published online by Cambridge University Press:
- 24 February 2020, pp. 1843-1850
- Print publication:
- 2020
-
- Article
- Export citation
-
Hot carrier based methods constitute a valuable approach for efficient and silicon compatible sub-bandgap photodetection. Although, hot electron excitation and transfer have been studied extensively on traditional materials such as Au and Ti, reports on alternative materials such as titanium nitride (TiN) are rare. Here, we perform hot hole photodetection measurements on a p-Si/metal thin film junction using Ti, Au and TiN. This material is of interest as it constitutes a refractory alternative to Au which is an important property for plasmonic applications where high field intensities can occur. In contrast to Au, a TiN/Si junction does not suffer from metal diffusion into the Si, which eases the integration with current Si-fabrication techniques. This work shows that a backside illuminated p-Si/TiN system can be used for efficient hot hole extraction in the IR, allowing for a responsivity of 1 mA/W at an excitation wavelength of 1250 nm and at zero bias. Via a comparison between TiN and other commonly used materials such as Au, the origin of this comparably high photoresponse can be traced back to be directly linked to a thin TiO2-x interfacial layer allowing for a distinct hot-hole transfer mechanism. Moreover, the fabrication of TiN nanodisk arrays is demonstrated which bears great promise to further boost the device efficiency.
Sleep problems in adolescence are prospectively linked to later depressive symptoms via the cortisol awakening response
- Kate Ryan Kuhlman, Jessica J. Chiang, Julienne E. Bower, Michael R. Irwin, Teresa E. Seeman, Heather E. McCreath, David M. Almeida, Ronald E. Dahl, Andrew J. Fuligni
-
- Journal:
- Development and Psychopathology / Volume 32 / Issue 3 / August 2020
- Published online by Cambridge University Press:
- 07 August 2019, pp. 997-1006
-
- Article
- Export citation
-
Sleep disturbance is a symptom of and a well-known risk factor for depression. Further, atypical functioning of the HPA axis has been linked to the pathogenesis of depression. The purpose of this study was to examine the role of adolescent HPA axis functioning in the link between adolescent sleep problems and later depressive symptoms. Methods: A sample of 157 17–18 year old adolescents (61.8% female) completed the Pittsburgh Sleep Quality Inventory (PSQI) and provided salivary cortisol samples throughout the day for three consecutive days. Two years later, adolescents reported their depressive symptoms via the Center for Epidemiological Studies Depression Scale (CES-D). Results: Individuals (age 17–18) with greater sleep disturbance reported greater depressive symptoms two years later (age 19–20). This association occurred through the indirect effect of sleep disturbance on the cortisol awakening response (CAR) (indirect effect = 0.14, 95%CI [.02 -.39]). Conclusions: One pathway through which sleep problems may lead to depressive symptoms is by up-regulating components of the body’s physiological stress response system that can be measured through the cortisol awakening response. Behavioral interventions that target sleep disturbance in adolescents may mitigate this neurobiological pathway to depression during this high-risk developmental phase.
3 - Favorite trajectories
-
- By Amy Bower, Department of Physical Oceanography, Woods Hole Oceanographic Institute, Woods Hole, Massachusetts, USA, Heather Furey, Department of Physical Oceanography, Woods Hole Oceanographic Institute, Woods Hole, Massachusetts, USA, Senya Grodsky, University of Maryland, College Park, Maryland, USA, Jim Carton, University of Maryland, College Park, Maryland, USA, Luca R. Centurioni, Scripps Institution of Oceanography, La Jolla, California, USA, Pearn P. Niiler, Scripps Institution of Oceanography, La Jolla, California, USA, Yoo Yin Kim, Scripps Institution of Oceanography La Jolla California USA, Dong-Kyu Lee, Busan National University, Busan, South Korea, Vitalii A. Sheremet, University of Rhode Island, Kingston, Rhode Island, USA, Newell Garfield, San Francisco State University, Tiburon, California, USA, Curtis A. Collins, Department of Oceanography, Naval Postgraduate School, Monterey, California, USA, Thomas A. Rago, Department of Oceanography, Naval Postgraduate School, Monterey, California, USA, Vassiliki Kourafalou, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida, USA, Elizabeth Williams, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida, USA, Thomas Lee, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida, USA, Matthias Lankhorst, Leibniz-Institut für Meereswissenschaften (IFM-GEOMAR), Kiel, Germany, Walter Zenk, Leibniz-Institut für Meereswissenschaften (IFM-GEOMAR), Kiel, Germany, Arthur J. Mariano, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida, USA, Edward H. Ryan, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida, USA, Pierre-Marie Poulain, Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS), Trieste, Italy, Hedinn Valdimarsson, Marine Research Institute, Reykjavik, Iceland, Svend-Aage Malmberg, Marine Research Institute, Reykjavik, Iceland
- Edited by Annalisa Griffa, University of Miami, A. D. Kirwan, Jr., University of Delaware, Arthur J. Mariano, University of Miami, Tamay Özgökmen, University of Miami, H. Thomas Rossby, University of Rhode Island
-
- Book:
- Lagrangian Analysis and Prediction of Coastal and Ocean Dynamics
- Published online:
- 07 September 2009
- Print publication:
- 10 May 2007, pp 68-88
-
- Chapter
- Export citation
-
Summary
In this chapter, a collection of “favorite trajectories” from various authors are presented.
While Lagrangian data analysis uses an extensive array of sophisticated tools, including classical statistics, dynamical system theory, stochastic modelling, assimilation techniques, and many others, visual inspection of individual trajectories still plays an important role, providing the first and often fundamental glimpse of the underlying dynamics. Often, for Lagrangian investigators, looking at trajectories gives the first intuition, then leading to the use of sophisticated and appropriate analysis. Trajectories tell the story of the journey of drifters and floats, and these stories are often complex and fascinating.
In the following sections, a number of investigators take us in the various world oceans, including Atlantic, Pacific and regional Seas, from the Poles to the Tropics, telling us the stories of their favorite trajectories and giving us their intuition and physical insights.
Mesoscale eddies in the Red Sea outflow region
In 2001–2002, 50 RAFOS floats were released at the core depth (∼ 650 m) of Red Sea Outflow Water (RSOW) in the Gulf of Aden (northwestern Indian Ocean) as part of the Red Sea Outflow Experiment (REDSOX). The objective was to determine how warm, saline RSOW spreads from its source at the southern end of Bab al Mandeb Strait to the open Indian Ocean. Our hypothesis was that either boundary undercurrents or submesoscale coherent vortices (SCVs like Meddies, but here called “Reddies”) were the main transport mechanisms for RSOW.