Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (59 results)

  • Thesis
  • Peer Reviewed
UC San Diego Electronic Theses and Dissertations (2006)
Visual neurons' spike trains represent a large variety of visual stimuli. The local contrast varies across natural scenes, and the absolute luminance changes that define visual features during low-contrast stimuli are much smaller than during high-contrast stimuli. The contrast of the scene can remain relatively stable for an extended period of timing, suggesting that it would be advantageous for the neuron to adjust its coding strategy to the stimulus contrast. However, the contrast can also rapidly change. If the neuron utilizes different coding strategies during different stimulus conditions, it is imperative that the neuron be able to recognize when the statistics of the stimulus have changed. We propose that neurons in the lateral geniculate nucleus (LGN) utilize nonlinear properties in order to encode visual information across a variety of stimulus conditions. In Chapters 2 and 3, we find that bursts and single spikes represent distinct stimuli, such that distinguishing between the bursts and single spikes provides information about the stimulus. Because bursts only occur following prolonged hyperpolarization, this suggests a means by which the neuron can encode the stimulus context: bursts may provide information about whether a stimulus is surprising given the recent stimulus history. In Chapter 4, we report the contrast normalization allows LGN neurons to encode information about stimuli across a wide range of local contrasts. Cells exhibiting strongest contrast normalization are best able to preserve information across stimuli. Furthermore, both the contrast normalization and the associated preservation of information could be reproduced by a non-adapting LGN model. In Chapter 5, we report other contrast-dependencies of the model, show transient changes in the model responses following a contrast change, and describe why the model is able to exhibit contrast normalization
    Cover page: Effects of non-linear processing on information transfer in the lateral geniculate nucleus
    • Article
    • Peer Reviewed
    UC Berkeley Previously Published Works (2023)
    We use the introduction of the Texas Top Ten Percent rule to estimate the effect of access to a selective college on graduation and earnings outcomes for two groups of students. For highly ranked students at more disadvantaged high schools, who gained access under the policy, college enrollment and graduation increased. Less highly ranked students at more advantaged schools, who tended to lose access, shifted toward less-selective colleges under the policy, but did not see declines in overall college enrollment, graduation, or earnings. The policy thus benefited students targeted for admission without evidence of adverse effects on displaced students.
      Cover page: Winners and Losers? The Effect of Gaining and Losing Access to Selective Colleges on Education and Labor Market Outcomes
      • Article
      • Peer Reviewed
      Western Journal of Emergency Medicine: Integrating Emergency Care with Population Health, Volume 18, Issue 5 (2017)

      Introduction: Deaths and injuries from all-terrain vehicle (ATV) crashes result in approximately700 deaths each year and more than 100,000 emergency department (ED) visits. Commonmisconceptions about ATV crashes are a significant barrier to injury prevention efforts, as is the lackof key information about where and how crashes occur. The purpose of this study was to determineATV crash patterns within a state, and to compare and contrast characteristics of these crashes as afunction of crash-site rurality.

      Methods: We performed descriptive, comparative, and regression analyses using a statewide off-roadvehicle crash and injury database (2002-2013). Comparisons were performed by rurality as defined usingthe Rural Urban Commuting Area (RUCA) coding system, and we used geographic information system(GIS) software to map crash patterns at the zip code and county levels.

      Results: ATV crashes occurred throughout the state; 46% occurred in urban and 54% in rural zip codeareas. Comparisons of rider and crash characteristics by rurality showed similarities by sex, age, seatingposition, on vs. off the road, and crash mechanism. Conversely, helmet use was significantly loweramong victims of isolated rural crashes as compared to other victims (p=0.004). Crashes in isolatedrural and small rural areas accounted for only 39% of all crashes but resulted in 62% of fatalities. In bothrural and urban areas, less than one-quarter of roadway injuries were traffic related. Relative crash ratesvaried by county, and unique patterns were observed for crashes involving youth and roadway riders.During the study period, 10% and 50% of all crashes occurred in 2% and 20% of the state’s counties,respectively.

      Conclusion: This study suggests that ATV crashes are a public health concern for both rural and urbancommunities. However, isolated rural ATV crash victims were less likely to be helmeted, and rural victimswere over-represented among fatalities. Traffic was not the major factor in roadway crashes in eitherrural or urban areas. Unique crash patterns for different riding populations suggest that injury preventionexperts and public policy makers should consider the potential impact of geographical location whendeveloping injury prevention interventions. [West J Emerg Med. 2017;18(5)913-922.]

        Cover page: Using Geospatial Mapping to Determine the Impact of All-Terrain Vehicle Crashes on Both Rural and Urban CommunitiesCreative Commons 'BY' version 4.0 license
        • Article
        • Peer Reviewed
        Proceedings of the Annual Meeting of the Cognitive Science Society, Volume 46 (2024)
        Large Language Models (LLMs), which match or exceed human performance on many linguistic tasks, are nonetheless commonly criticized for not “understanding” language. These critiques are hard to evaluate because they conflate “understanding” with reasoning and common sense—abilities that, in human minds, are dissociated from language processing per se. Here, we instead focus on a form of understanding that is tightly linked to language: mapping sentence structure onto an event description of “who did what to whom” (thematic roles). Whereas LLMs can be directly trained to solve to this task, we asked whether they naturally learn to extract such information during their regular, unsupervised training on word prediction. In two experiments, we evaluated sentence representations in two commonly used LLMs—BERT and GPT-2. Experiment 1 tested hidden representations distributed across all hidden units, and found an unexpected pattern: sentence pairs that had opposite (reversed) agent and patient, but shared syntax, were represented as more similar than pairs that shared the same agent and same patient, but differed in syntax. In contrast, human similarity judgments were driven by thematic role assignment. Experiment 2 asked whether thematic role information was localized to a subset of units and/or to attention heads. We found little evidence that this information was available in hidden units (with one exception). However, we found attention heads that reflected thematic roles independent of syntax. Therefore, some components within LLMs capture thematic roles, but such information exerts a much weaker influence on their sentence representations compared to its influence on human judgments.
          Cover page: Do Large language Models know who did what to whom?Creative Commons 'BY' version 4.0 license
          • Article
          • Peer Reviewed
          Faculty Publications (2008)
          The Amazon Basin is crucial to global circulatory and carbon patterns due to the large areal extent and large flux magnitude. Biogeophysical models have had difficulty reproducing the annual cycle of net ecosystem exchange (NEE) of carbon in some regions of the Amazon, generally simulating uptake during the wet season and efflux during seasonal drought. In reality, the opposite occurs. Observational and modeling studies have identified several mechanisms that explain the observed annual cycle, including: (1) deep soil columns that can store large water amount, (2) the ability of deep roots to access moisture at depth when near-surface soil dries during annual drought, (3) movement of water in the soil via hydraulic redistribution, allowing for more efficient uptake of water during the wet season, and moistening of near-surface soil during the annual drought, and (4) photosynthetic response to elevated light levels as cloudiness decreases during the dry season. We incorporate these mechanisms into the third version of the Simple Biosphere model (SiB3) both singly and collectively, and confront the results with observations. For the forest to maintain function through seasonal drought, there must be sufficient water storage in the soil to sustain transpiration through the dry season in addition to the ability of the roots to access the stored water. We find that individually, none of these mechanisms by themselves produces a simulation of the annual cycle of NEE that matches the observed. When these mechanisms are combined into the model, NEE follows the general trend of the observations, showing efflux during the wet season and uptake during seasonal drought.
            Cover page: Seasonal drought stress in the Amazon: Reconciling models and observationsCreative Commons 'BY' version 3.0 license
            • Article
            • Peer Reviewed
            UC San Diego Previously Published Works (2022)
            INTRODUCTION: Emotional processes play a role in both suicide risk and eating disorders (EDs), which are often comorbid. However, limited research has explored how emotional processes relate to suicide risk in EDs and the prognostic value of suicide risk for ED treatment. Thus, the current study examined associations between emotion dysregulation and reactivity with suicide risk in patients with EDs, and determined if suicide risk predicts ED treatment outcomes. METHODS: Participants (n = 201) were adults in an ED partial hospitalization program who completed measures at admission, 1-month post-admission, and discharge. RESULTS: When controlling for depressive symptoms, limited access to adaptive emotion regulation strategies, difficulties engaging in goal-oriented behaviors, and engaging in impulsive behavior when experiencing negative emotions (i.e., emotion dysregulation) were associated with suicide attempt frequency. Depressive symptoms were associated with suicide risk severity, while emotion dysregulation and reactivity were not. Importantly, patients with elevated suicide risk at admission improved comparably to other risk categories across treatment, despite presenting with greater ED symptoms at admission. CONCLUSION: Emotion dysregulation and depression are salient factors when examining suicide risk in patients with EDs. Suicide risk and attempt history may not negatively impact ED treatment outcomes when using emotion-focused treatment.
              Cover page: Associations of suicide risk with emotional reactivity, dysregulation, and eating disorder treatment outcomes.
              • Article
              • Peer Reviewed
              Western Journal of Emergency Medicine: Integrating Emergency Care with Population Health, Volume 16, Issue 5 (2015)
              Introduction: National studies of largely urban populations showed increased risk of traumatic death among uninsured patients, as compared to those insured. No similar studies have been done for major trauma centers serving rural states. Methods: We performed retrospective analyses using trauma registry records from adult, non-burn patients admitted to a single American College of Surgeons-certified Level 1 trauma center in a rural state (2003-2010, n=13,680) and National Trauma Data Bank (NTDB) registry records (2002-2008, n=380,182). Risk of traumatic death was estimated using multivariable logistic regression analysis. Results: We found that 9% of trauma center patients and 27% of NTDB patients were uninsured. Overall mortality was similar for both (~4.5%). After controlling for covariates, uninsured trauma center patients were almost five times more likely to die and uninsured NTDB patients were 75% more likely to die than commercially insured patients. The risk of death among Medicaid patients was not significantly different from the commercially insured for either dataset. Conclusion: Our results suggest that even with an inclusive statewide trauma system and an emergency department that does not triage by payer status, uninsured patients presenting to the trauma center were at increased risk of traumatic death relative to patients with commercial insurance.
                Cover page: Not Just an Urban Phenomenon: Uninsured Rural Trauma Patients at Increased Risk for MortalityCreative Commons 'BY' version 4.0 license
                • Article
                • Peer Reviewed
                UC Irvine Previously Published Works (2001)
                The nucleoporins Nup60p, Nup2p, and Nup1p form part of the nuclear basket structure of the Saccharomyces cerevisiae nuclear pore complex (NPC). Here, we show that these necleoporins can be isolated from yeast extracts by affinity chromatography on karyopherin Kap95p-coated beads. To characterize Nup60p further, Nup60p-coated beads were used to capture its interacting proteins from extracts. We find that Nup60p binds to Nup2p and serves as a docking site for Kap95p-Kap60p heterodimers and Kap123p. Nup60p also binds Gsp1p-GTP and its guanine nucleotide exchange factor Prp20p, and functions as a Gsp1p guanine nucleotide dissociation inhibitor by reducing the activity of Prp20p. Yeast lacking Nup60p exhibit minor defects in nuclear export of Kap60p, nuclear import of Kap95p-Kap60p-dependent cargoes, and diffusion of small proteins across the NPC. Yeast lacking Nup60p also fail to anchor Nup2p at the NPC, resulting in the mislocalization of Nup2p to the nucleoplasm and cytoplasm. Purified Nup60p and Nup2p bind each other directly, but the stability of the complex is compromised when Kap60p binds Nup2p. Gsp1p-GTP enhances by 10-fold the affinity between Nup60p and Nup2p, and restores binding of Nup2p-Kap60p complexes to Nup60p. The results suggest a dynamic interaction, controlled by the nucleoplasmic concentration of Gsp1p-GTP, between Nup60p and Nup2p at the NPC.
                  Cover page: The nucleoporin Nup60p functions as a Gsp1p–GTP-sensitive tether for Nup2p at the nuclear pore complex
                  • Article
                  • Peer Reviewed
                  LBL Publications (2017)
                  The interaction of land with the atmosphere is sensitive to soil moisture (W). Evapotranspiration (ET) reacts to soil moisture in a nonlinear way, f(W), as soils dry from saturation to wilt point. This nonlinear behavior and the fact that soil moisture varies on scales as small as 1–10 m in nature, while numerical general circulation models (GCMs) have grid cell sizes on the order of 1 to 100s of kilometers, makes the calculation of grid cell-average ET problematic. It is impractical to simulate the land in GCMs on the small scales seen in nature, so techniques have been developed to represent subgrid scale heterogeneity, including: (1) statistical-dynamical representations of grid subelements of varying wetness, (2) relaxation of f(W), (3) moderating f(W) with approximations of catchment hydrology, (4) “tiling” the landscape into vegetation types, and (5) hyperresolution. Here we present an alternative method for representing subgrid variability in W, one proven in a conceptual framework where landscape-scale W is represented as a series of “Bins” of increasing wetness from dry to saturated. The grid cell-level f(W) is defined by the integral of the fractional area of the wetness bins and the value of f(W) associated with each. This approach accounts for the spatiotemporal dynamics of W. We implemented this approach in the SiB3 land surface parameterization and then evaluated its performance against a control, which assumes a horizontally uniform field of W. We demonstrate that the Bins method, with a physical basis, attenuates unrealistic jumps in model state and ET seen in the control runs.
                    Cover page: Closing the scale gap between land surface parameterizations and GCMs with a new scheme, SiB3‐Bins
                    • Article
                    • Peer Reviewed
                    UC San Francisco Previously Published Works (2012)
                    Electron microscopy (EM) and atomic force microscopy (AFM) techniques have long been used to characterize collagen fibril ordering and alignment in connective tissues. These techniques, however, are unable to map collagen fibril polarity, i.e., the polar orientation that is directed from the amine to the carboxyl termini. Using a voltage modulated AFM-based technique called piezoresponse force microscopy (PFM), we show it is possible to visualize both the alignment of collagen fibrils within a tissue and the polar orientation of the fibrils with minimal sample preparation. We demonstrate the technique on rat tail tendon and porcine eye tissues in ambient conditions. In each sample, fibrils are arranged into domains whereby neighboring domains exhibit opposite polarizations, which in some cases extend to the individual fibrillar level. Uniform polarity has not been observed in any of the tissues studied. Evidence of anti-parallel ordering of the amine to carboxyl polarity in bundles of fibrils or in individual fibrils is found in all tissues, which has relevance for understanding mechanical and biofunctional properties and the formation of connective tissues. The technique can be applied to any biological material containing piezoelectric biopolymers or polysaccharides.
                      Cover page: Visualizing molecular polar order in tissues via electromechanical coupling
                      Top