scholarly journals Sensitivity as a Complexity Measure for Sequence Classification Tasks

2021 ◽  
Vol 9 ◽  
pp. 891-908
Author(s):  
Michael Hahn ◽  
Dan Jurafsky ◽  
Richard Futrell
Keyword(s):  
Boolean Function ◽  
Text Classification ◽  
Input Sequence ◽  
High Sensitivity ◽  
Sequence Classification ◽  
Classification Methods ◽  
Sensitivity Functions ◽  
Challenging Tasks ◽  
Classification Tasks ◽  
Low Sensitivity

Abstract We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint subsets of the input sequence that can each be individually changed to change the output. We argue that standard sequence classification methods are biased towards learning low-sensitivity functions, so that tasks requiring high sensitivity are more difficult. To that end, we show analytically that simple lexical classifiers can only express functions of bounded sensitivity, and we show empirically that low-sensitivity functions are easier to learn for LSTMs. We then estimate sensitivity on 15 NLP tasks, finding that sensitivity is higher on challenging tasks collected in GLUE than on simple text classification tasks, and that sensitivity predicts the performance both of simple lexical classifiers and of vanilla BiLSTMs without pretrained contextualized embeddings. Within a task, sensitivity predicts which inputs are hard for such simple models. Our results suggest that the success of massively pretrained contextual representations stems in part because they provide representations from which information can be extracted by low-sensitivity decoders.

scholarly journals Concentration Bounds for High Sensitivity Functions Through Differential Privacy

2019 ◽  
Vol 9 (1) ◽  
Author(s):  
Uri Stemmer ◽  
Kobbi Nissim
Keyword(s):  
Data Analysis ◽  
Differential Privacy ◽  
High Sensitivity ◽  
Sensitivity Function ◽  
Mathematical Tool ◽  
Sensitivity Functions ◽  
Chernoff Bound ◽  
Concentration Bounds ◽  
Adaptive Data Analysis ◽  
Low Sensitivity

A new line of work demonstrates how differential privacy can be used as a mathematical tool for guaranteeing generalization in adaptive data analysis. Specifically, if a differentially private analysis is applied on a sample S of i.i.d. examples to select a low-sensitivity function f, then w.h.p. f(S) is close to its expectation, even though f is being chosen adaptively, i.e., based on the data. Very recently, Steinke and Ullman observed that these generalization guarantees can be used for proving concentration bounds in the non-adaptive setting, where the low-sensitivity function is fixed beforehand. In particular, they obtain alternative proofs for classical concentration bounds for low-sensitivity functions, such as the Chernoff bound and McDiarmid's Inequality. In this work, we extend this connection between differential privacy and concentration bounds, and show that differential privacy can be used to prove concentration of high-sensitivity functions.

scholarly journals Second harmonic generation correlation spectroscopy for characterizing translationally diffusing protein nanocrystals

2016 ◽  
Vol 72 (7) ◽  
pp. 849-859
Author(s):  
Ximeng Y. Dow ◽  
Christopher M. Dettmar ◽  
Emma L. DeWalt ◽  
Justin A. Newman ◽  
Alexander R. Dow ◽  
...  
Keyword(s):  
Second Harmonic Generation ◽  
Harmonic Generation ◽  
Unit Volume ◽  
Second Harmonic ◽  
Incident Beam ◽  
High Sensitivity ◽  
Correlation Spectroscopy ◽  
Optical Process ◽  
Active Nanoparticles ◽  
Low Sensitivity

Second harmonic generation correlation spectroscopy (SHG-CS) is demonstrated as a new approach to protein nanocrystal characterization. A novel line-scanning approach was performed to enable autocorrelation analysis without sample damage from the intense incident beam. An analytical model for autocorrelation was developed, which includes a correction for the optical scattering forces arising when focusing intense, infrared beams. SHG-CS was applied to the analysis of BaTiO3nanoparticles ranging from 200 to ∼500 nm and of photosystem I nanocrystals. A size distribution was recovered for each sample and compared with the size histogram measured by scanning electron microscopy (SEM). Good agreement was observed between the two independent measurements. The intrinsic selectivity of the second-order nonlinear optical process provides SHG-CS with the ability to distinguish well ordered nanocrystals from conglomerates and amorphous aggregates. Combining the recovered distribution of particle diameters with the histogram of measured SHG intensities provides the inherent hyperpolarizability per unit volume of the SHG-active nanoparticles. Simulations suggest that the SHG activity per unit volume is likely to exhibit relatively low sensitivity to the subtle distortions within the lattice that contribute to resolution loss in X-ray diffraction, but high sensitivity to the presence of multi-domain crystals.

Influence of Irrational Beliefs and Sensitivity to a Person on Future Psychology Students’ Propensity to Coercion or Nonviolence

2021 ◽  
pp. 143-165
Author(s):  
V.G. Maralov ◽  
◽  
V.A. Sitarov ◽  
Keyword(s):  
Irrational Beliefs ◽  
High Sensitivity ◽  
Daily Practice ◽  
Psychology Students ◽  
State Universities ◽  
Low Level ◽  
Level Of Understanding ◽  
Low Sensitivity ◽  
High Level ◽  
Dependent Type

The relevance of the problem is due to the importance of identifying factors that determine the propensity of students to coercion or nonviolence, creating psychological and pedagogical conditions for the formation of the socionomic sphere of nonviolent competencies for future specialists at universities. The theoretical basis of the study was the position of nonviolence as a daily practice of interaction, by which we understand the ability of a person to choose from a number of possible alternatives that carry the least charge of coercion. The aim of the work was to study the influence of irrational beliefs and sensitivity to a person (interest, empathy, understanding and assistance) on the students’ tendency to coercion, manipulation, non-violence and non-interference in the processes of interaction with people. The hypothesis was tested that the tendency of students to coercion, manipulation, and noninterference will be due to expressed irrational beliefs and low level of sensitivity to a person and the tendency to non-violence will be explained by the absence of irrational beliefs and a high level of sensitivity to a person. The study involved 125 students of pedagogical and psychological faculties of the Moscow Humanitarian and Cherepovets State universities. The authors used questionnaires to identify the positions of interaction among students and sensitivity to a person, as well as a list of irrational beliefs proposed by A. Beck and A. Freeman. It is established that the tendency to both coercion and manipulation are determined by the beliefs of anti-social type and low sensitivity to the person. The tendency to manipulate the narcissistic beliefs, high interest in people and understanding them, at the same time the tendency to non-violence and non-interference are determined by beliefs of avoidant and dependent types with a low level of understanding people. And a tendency to non-interference is determined by beliefs of dependent type with unexpressed orientation on helping. The tendency to nonviolence is determined by the high sensitivity to a person and the absence of irrational beliefs of antisocial, passive-aggressive and narcissistic types. As a result, the conclusion is made about the need to form purposefully the ability to nonviolent interaction among students, which should include the work on awareness and overcoming irrational beliefs and the development of sensitivity to a person. The obtained results can be used in practical work with students on the formation of their nonviolent competencies.

scholarly journals Predicting Response to Total Neoadjuvant Treatment (TNT) in Locally Advanced Rectal Cancer Based on Multiparametric Magnetic Resonance Imaging: A Retrospective Study

2020 ◽  
Author(s):  
Ganlu Ouyang ◽  
Xibiao Yang ◽  
Xiangbing Deng ◽  
Wenjian Meng ◽  
Yongyang Yu ◽  
...  
Keyword(s):  
Magnetic Resonance Imaging ◽  
Rectal Cancer ◽  
Lymph Node ◽  
Locally Advanced ◽  
Neoadjuvant Treatment ◽  
High Sensitivity ◽  
Advanced Rectal Cancer ◽  
Locally Advanced Rectal Cancer ◽  
Resonance Imaging ◽  
Low Sensitivity

Abstract Purpose: To investigate the potential value of magnetic resonance imaging (MRI) in predicting response relevance to total neoadjuvant treatment (TNT) in locally advanced rectal cancer.Methods: We analyzed MRI of 71 patients underwent TNT from 2015 to 2017 retrospectively. We categorized the response of TNT as CR (complete response) and non-CR, and high, moderate and low sensitivity. Logistic regression analysis was used to identify the best predictors of response. Diagnostic performance was assessed using receiver - operating characteristic curve analysis.Results: Post–ICT (induction chemotherapy) ∆TL (tumor length), post-CRT (concurrent chemoradiotherapy) ∆LNN (the numbers of lymph node metastases), post–CCT (consolidation chemotherapy) ∆SDWI (maximum cross-sectional area of tumor on diffusion-weighted imaging), post-CCT ADCT (the mean apparent diffusion coefficient values of tumor) and post-CCT ∆LNV (volume of lymph node) were the best CR predictors. Post-CRT EMVI (extramural vascular invasion) and post-CCT ∆ST2 (S on T2-weight) were the best significant factors for high sensitivity. Conclusions: Post-ICT ∆TL and post-CRT EMVI may an early predictor of CR and high sensitivity to TNT, respectively. The grouping scheme of CR and non–CR was more suitable for predicting response by MRI parameters than high, moderate and low sensitivity.Trial registration: retrospectively registered

scholarly journals The Effect of Oral Tactile Sensitivity on Texture Perception and Mastication Behavior in Humans

2018 ◽  
Author(s):  
Grace E. Shupe ◽  
Arran Wilson ◽  
Curtis R. Luckett
Keyword(s):  
High Sensitivity ◽  
Tactile Feedback ◽  
Interindividual Variation ◽  
Texture Perception ◽  
Tactile Acuity ◽  
Force Sensitivity ◽  
Significant Difference ◽  
Low Sensitivity ◽  
Oral Sensitivity ◽  
The Brain

AbstractMastication behavior is a notable source of interindividual variation in texture perception and could be linked to oral sensitivity. As oral sensitivity declines so does the amount of tactile feedback relayed to the brain, resulting in less effective manipulation or food and a reduced ability to discriminate differences. To address these hypotheses, we measured masticatory behavior and related this to texture discrimination and oral sensitivity. The study was performed on 41 participants in two groups, with high (n = 20) or low (n=21) sensitivity. Oral sensitivity was measured using a battery of tests that included: oral stereognosis, lingual tactile acuity, and bite force sensitivity. Sensitivity to texture changes was measured using a series of triangle tests with confectionaries of different hardness, with masticatory patterns and behaviors being video recorded and analyzed using jaw tracking software. Overall, there was no significant difference between high and low sensitivity participants and their ability to distinguish texture changes. But, there were significantly different trends found between the groups based on their masticatory behaviors including chewing pattern and overall number of chews. But, it was found that multiple masticatory behaviors were being modulated by oral sensitivity, including overall chewing cycles used (p < 0.0001). More, specifically those in the high sensitivity group used more stochastic chewing movements, while those in the low sensitivity group were found to use crescent-shaped chewing cycles. It was also noted that in the high sensitivity group the jaw moved further distances (p < 0.0001) in all phases and moved at a higher velocity when opening (p < 0.0001) but not when closing, when compared to the low sensitivity group. These results help bolster evidence that mastication and oral sensitivity are related.

Sign in / Sign up

Export Citation Format

Share Document