Search

Recent advances in deep learning have led to interest in training deep learning models on longitudinal healthcare records to predict a range of medical events, with models demonstrating high predictive performance. Predictive performance is necessary but insufficient, however, with explanations and reasoning from models required to convince clinicians for sustained use. Rigorous evaluation of explainability is often missing, as comparisons between models (traditional versus deep) and various explainability methods have not been well-studied. Furthermore, ground truths needed to evaluate explainability can be highly subjective depending on the clinician's perspective. Our work is one of the first to evaluate explainability performance between and within traditional (XGBoost) and deep learning (LSTM with Attention) models on both a global and individual per-prediction level on longitudinal healthcare data. We compared explainability using three popular methods: 1) SHapley Additive exPlanations (SHAP), 2) Layer-Wise Relevance Propagation (LRP), and 3) Attention. These implementations were applied on synthetically generated datasets with designed ground-truths and a real-world medicare claims dataset. We showed that overall, LSTMs with SHAP or LRP provides superior explainability compared to XGBoost on both the global and local level, while LSTM with dot-product attention failed to produce reasonable ones. With the explosion of the volume of healthcare data and deep learning progress, the need to evaluate explainability will be pivotal towards successful adoption of deep learning models in healthcare settings., Comment: 21 pages, 10 figures

The XGBoost method has many advantages and is especially suitable for statistical analysis of big data, but its loss function is limited to convex functions. In many specific applications, a nonconvex loss function would be preferable. In this paper, I propose a generalized XGBoost method, which requires weaker loss function constraint and involves more general loss functions, including convex loss functions and some non-convex loss functions. Furthermore, this generalized XGBoost method is extended to multivariate loss function to form a more generalized XGBoost method. This method is a multiobjective parameter regularized tree boosting method, which can model multiple parameters in most of the frequently-used parametric probability distributions to be fitted by predictor variables. Meanwhile, the related algorithms and some examples in non-life insurance pricing are given., Comment: The second version,Fixed some language expressions

Two-dimensional (2D) hybrid perovskite sandwiched between two long-chain organic layers is an emerging class of low-cost semiconductor materials with unique optical properties and improved moisture stability. Unlike conventional semiconductors, ion migration in perovskite is a unique phenomenon possibly responsible for long carrier lifetime, current-voltage hysteresis, and low-frequency giant dielectric response. While there are many studies of ion migration in bulk hybrid perovskite, not much is known for its 2D counterparts, especially for ion migration induced by light excitation. Here, we construct an exfoliated 2D perovskite/carbon nanotube (CNT) heterostructure field effect transistor (FET), not only to demonstrate its potential in photomemory applications, but also to study the light induced ion migration mechanisms. We show that the FET I-V characteristic curve can be regulated by light and shows two opposite trends under different CNT oxygen doping conditions. Our temperature-dependent study indicates that the change in the I-V curve is probably caused by ion redistribution in the 2D hybrid perovskite. The first principle calculation shows the reduction of the migration barrier of I vacancy under light excitation. The device simulation shows that the increase of 2D hybrid perovskite dielectric constant (enabled by the increased ion migration) can change the I-V curve in the trends observed experimentally. Finally, the so synthesized FET shows the multilevel photomemory function. Our work shows that not only we could understand the unique ion migration behavior in 2D hybrid perovskite, it might also be used for many future memory function related applications not realizable in traditional semiconductors.

Two-dimensional (2D) hybrid perovskite sandwiched between two long-chain organic layers is an emerging class of low-cost semiconductor materials with unique optical properties and improved moisture stability. Unlike conventional semiconductors, ion migration in perovskite is a unique phenomenon possibly responsible for long carrier lifetime, current-voltage hysteresis, and low-frequency giant dielectric response. While there are many studies of ion migration in bulk hybrid perovskite, not much is known for its 2D counterparts, especially for ion migration induced by light excitation. Here, we construct an exfoliated 2D perovskite/carbon nanotube (CNT) heterostructure field effect transistor (FET), not only to demonstrate its potential in photomemory applications, but also to study the light induced ion migration mechanisms. We show that the FET I-V characteristic curve can be regulated by light and shows two opposite trends under different CNT oxygen doping conditions. Our temperature-dependent study indicates that the change in the I-V curve is probably caused by ion redistribution in the 2D hybrid perovskite. The first principle calculation shows the reduction of the migration barrier of I vacancy under light excitation. The device simulation shows that the increase of 2D hybrid perovskite dielectric constant (enabled by the increased ion migration) can change the I-V curve in the trends observed experimentally. Finally, the so synthesized FET shows the multilevel photomemory function. Our work shows that not only we could understand the unique ion migration behavior in 2D hybrid perovskite, it might also be used for many future memory function related applications not realizable in traditional semiconductors.

In a k-party communication problem, the k players with inputs x_1, x_2, ..., x_k, respectively, want to evaluate a function f(x_1, x_2, ..., x_k) using as little communication as possible. We consider the message-passing model, in which the inputs are partitioned in an arbitrary, possibly worst-case manner, among a smaller number t of players (t= 2/3, in a strict turnstile stream. Our result matches the best known upper bound when epsilon >= 1/polylog(mM). It also improves on the prior Omega({epsilon}^{-2}log(mM)) lower bound and separates the complexity of approximating L_0 from approximating the p-norm L_p for p bounded away from 0, since the latter has an O(epsilon^{-2}log(mM)) bit upper bound.

We consider algorithms with access to an unknown matrix M in F^{n x d} via matrix-vector products, namely, the algorithm chooses vectors v^1, ..., v^q, and observes Mv^1, ..., Mv^q. Here the v^i can be randomized as well as chosen adaptively as a function of Mv^1, ..., Mv^{i-1}. Motivated by applications of sketching in distributed computation, linear algebra, and streaming models, as well as connections to areas such as communication complexity and property testing, we initiate the study of the number q of queries needed to solve various fundamental problems. We study problems in three broad categories, including linear algebra, statistics problems, and graph problems. For example, we consider the number of queries required to approximate the rank, trace, maximum eigenvalue, and norms of a matrix M; to compute the AND/OR/Parity of each column or row of M, to decide whether there are identical columns or rows in M or whether M is symmetric, diagonal, or unitary; or to compute whether a graph defined by M is connected or triangle-free. We also show separations for algorithms that are allowed to obtain matrix-vector products only by querying vectors on the right, versus algorithms that can query vectors on both the left and the right. We also show separations depending on the underlying field the matrix-vector product occurs in. For graph problems, we show separations depending on the form of the matrix (bipartite adjacency versus signed edge-vertex incidence matrix) to represent the graph. Surprisingly, this fundamental model does not appear to have been studied on its own, and we believe a thorough investigation of problems in this model would be beneficial to a number of different application areas.

A circuit C compresses a function f:{0,1}^n -> {0,1}^m if given an input x in {0,1}^n the circuit C can shrink x to a shorter l-bit string x' such that later, a computationally-unbounded solver D will be able to compute f(x) based on x'. In this paper we study the existence of functions which are incompressible by circuits of some fixed polynomial size s=n^c. Motivated by cryptographic applications, we focus on average-case (l,epsilon) incompressibility, which guarantees that on a random input x in {0,1}^n, for every size s circuit C:{0,1}^n -> {0,1}^l and any unbounded solver D, the success probability Pr_x[D(C(x))=f(x)] is upper-bounded by 2^(-m)+epsilon. While this notion of incompressibility appeared in several works (e.g., Dubrov and Ishai, STOC 06), so far no explicit constructions of efficiently computable incompressible functions were known. In this work we present the following results: 1. Assuming that E is hard for exponential size nondeterministic circuits, we construct a polynomial time computable boolean function f:{0,1}^n -> {0,1} which is incompressible by size n^c circuits with communication l=(1-o(1)) * n and error epsilon=n^(-c). Our technique generalizes to the case of PRGs against nonboolean circuits, improving and simplifying the previous construction of Shaltiel and Artemenko (STOC 14). 2. We show that it is possible to achieve negligible error parameter epsilon=n^(-omega(1)) for nonboolean functions. Specifically, assuming that E is hard for exponential size Sigma_3-circuits, we construct a nonboolean function f:{0,1}^n -> {0,1}^m which is incompressible by size n^c circuits with l=Omega(n) and extremely small epsilon=n^(-c) * 2^(-m). Our construction combines the techniques of Trevisan and Vadhan (FOCS 00) with a new notion of relative error deterministic extractor which may be of independent interest. 3. We show that the task of constructing an incompressible boolean function f:{0,1}^n -> {0,1} with negligible error parameter epsilon cannot

The left-right twin Higgs model contains a new vector-like heavy top quark, which mixes with the SM-like top quark. In this work, we studied the single vector-like top partner production via process $e^{-}\gamma \rightarrow \nu_{e}\bar{T}b$ at the International Linear Collider. We calculated the production cross section at tree level and displayed the relevant differential distributions. The result shows that there will be 125 events produced each year with $\sqrt{s}$=2TeV and the integrated luminosity $\mathcal L_{int}\simeq 500fb^{-1}$, and the b-quark tagging and the relevant missing energy $\met$ cut will be helpful to detect this new effect., Comment: Submitted to Chinese Physics C

The left-right twin Higgs model contains a new vector-like heavy top quark, which mixes with the SM-like top quark. In this work, we studied the single vector-like top partner production via process $e^{-}\gamma \rightarrow \nu_{e}\bar{T}b$ at the International Linear Collider. We calculated the production cross section at tree level and displayed the relevant differential distributions. The result shows that there will be 125 events produced each year with $\sqrt{s}$=2TeV and the integrated luminosity $\mathcal L_{int}\simeq 500fb^{-1}$, and the b-quark tagging and the relevant missing energy $\met$ cut will be helpful to detect this new effect., Comment: Submitted to Chinese Physics C