BZPEER

preprint2021arXiv

The Analysis from Nonlinear Distance Metric to Kernel-based Drug Prescription Prediction System

Distance metrics and their nonlinear variant play a crucial role in machine learning based real-world problem solving. We demonstrated how Euclidean and cosine distance measures differ not only theoretically but also in real-world medical application, namely, outcome prediction of drug prescription. Euclidean distance exhibits favorable properties in the local geometry problem. To this regard, Euclidean distance can be applied under short-term disease with low-variation outcome observation. Moreover, when presenting to highly variant chronic disease, it is preferable to use cosine distance. These different geometric properties lead to different submanifolds in the original embedded space, and hence, to different optimizing nonlinear kernel embedding frameworks. We first established the geometric properties that we needed in these frameworks. From these properties interpreted their differences in certain perspectives. Our evaluation on real-world, large-scale electronic health records and embedding space visualization empirically validated our approach.

preprint2020arXiv

Cross-Global Attention Graph Kernel Network Prediction of Drug Prescription

We present an end-to-end, interpretable, deep-learning architecture to learn a graph kernel that predicts the outcome of chronic disease drug prescription. This is achieved through a deep metric learning collaborative with a Support Vector Machine objective using a graphical representation of Electronic Health Records. We formulate the predictive model as a binary graph classification problem with an adaptive learned graph kernel through novel cross-global attention node matching between patient graphs, simultaneously computing on multiple graphs without training pair or triplet generation. Results using the Taiwanese National Health Insurance Research Database demonstrate that our approach outperforms current start-of-the-art models both in terms of accuracy and interpretability.

preprint2016arXiv

Gradient Estimates via Rearrangements for Solutions of Some Schrödinger Equations

In this article, by applying the well known method for dealing with $p$ -Laplace type elliptic boundary value problems, the authors establish a sharp estimate for the decreasing rearrangement of the gradient of solutions to the Dirichlet and the Neumann boundary value problems of a class of Schrödinger equations, under the weak regularity assumption on the boundary of domains. As applications, gradient estimates of these solutions in Lebesgue spaces and Lorentz spaces are obtained.

preprint2016arXiv

Littlewood-Paley Characterizations of Hajłasz-Sobolev and Triebel-Lizorkin Spaces via Averages on Balls

Let $p\in(1,\infty)$ and $q\in[1,\infty)$ . In this article, the authors characterize the Triebel-Lizorkin space ${F}^α_{p,q}(\mathbb{R}^n)$ with smoothness order $α\in(0,2)$ via the Lusin-area function and the $g_λ^*$ -function in terms of difference between $f(x)$ and its average $B_tf(x):=\frac1{|B(x,t)|}\int_{B(x,t)}f(y)\,dy$ over a ball $B(x,t)$ centered at $x\in\mathbb{R}^n$ with radius $t\in(0,1)$ . As an application, the authors obtain a series of characterizations of $F^α_{p,\infty}(\mathbb{R}^n)$ via pointwise inequalities, involving ball averages, in spirit close to Hajłasz gradients, here an interesting phenomena naturally appears that, in the end-point case when $α=2$ , these pointwise inequalities characterize the Triebel-Lizorkin spaces $F^2_{p,2}(\mathbb{R}^n)$ , while not $F^2_{p,\infty}(\mathbb{R}^n)$ . In particular, some new pointwise characterizations of Hajłasz-Sobolev spaces via ball averages are obtained. Since these new characterizations only use ball averages, they can be used as starting points for developing a theory of Triebel-Lizorkin spaces with smoothness orders not less than $1$ on spaces of homogeneous type.

preprint2016arXiv

On Li-Yau gradient estimate for sum of squares of vector fields up to higher step

In this paper, we generalize the Cao-Yau's gradient estimate for the sum of squares of vector fields up to higher step under assumption of the generalized curvature-dimension inequality. With its applications, by deriving a curvature-dimension inequality, we are able to obtain the Li-Yau gradient estimate for the CR heat equation in a closed pseudohermitian manifold of nonvanishing torsion tensors. As consequences, we obtain the Harnack inequality and upper bound estimate for the CR heat kernel.

preprint2015arXiv

On the Hodge-type decomposition and cohomolgy groups of $k$ -Cauchy-Fueter complexes over domains in the quaternionic space

The $k$ -Cauchy-Fueter operator $D_0^{(k) }$ on one dimensional quaternionic space $\mathbb{H}$ is the Euclidean version of helicity $\frac k 2$ massless field operator on the Minkowski space in physics. The $k$ -Cauchy-Fueter equation for $k\geq 2$ is overdetermined and its compatibility condition is given by the $k$ -Cauchy-Fueter complex. In quaternionic analysis, these complexes play the role of Dolbeault complex in several complex variables. We prove that a natural boundary value problem associated to this complex is regular. Then by using the theory of regular boundary value problems, we show the Hodge-type orthogonal decomposition, and the fact that the non-homogeneous $k$ -Cauchy-Fueter equation $D_0^{(k) } u=f$ on a smooth domain $Ω$ in $\mathbb{H}$ is solvable if and only if $f$ satisfies the compatibility condition and is orthogonal to the set $\mathscr H^1_{ (k) }(Ω)$ of Hodge-type elements. This set is isomorphic to the first cohomology group of the $k$ -Cauchy-Fueter complex over $Ω$ , which is finite dimensional, while the second cohomology group is always trivial.

preprint2014arXiv

Riesz Transform Characterizations of Musielak-Orlicz-Hardy Spaces

Let $φ$ be a Musielak-Orlicz function satisfying that, for any $(x,\,t)\in\mathbb{R}^n\times(0,\,\infty)$ , $φ(\cdot,\,t)$ belongs to the Muckenhoupt weight class $A_\infty (\mathbb{R}^n)$ with the critical weight exponent $q(φ)\in[1,\,\infty)$ and $φ(x,\,\cdot)$ is an Orlicz function with $0<i(φ)\le I(φ)\le 1$ which are, respectively, its critical lower type and upper type. In this article, the authors establish the Riesz transform characterizations of the Musielak-Orlicz-Hardy spaces $H_φ(\mathbb{R}^n)$ which are generalizations of weighted Hardy spaces and Orlicz-Hardy spaces. Precisely, the authors characterize $H_φ(\mathbb{R}^n)$ via all the first order Riesz transforms when $\frac{i(φ)}{q(φ)}>\frac{n-1}{n}$ , and via all the Riesz transforms with the order not more than $m\in\mathbb{N}$ when $\frac{i(φ)}{q(φ)}>\frac{n-1}{n+m-1}$ . Moreover, the authors also establish the Riesz transform characterizations of $H_φ(\mathbb{R}^n)$ , respectively, by means of the higher order Riesz transforms defined via the homogenous harmonic polynomials or the odd order Riesz transforms. Even if when $φ(x,t):=tw(x)$ for all $x\in{\mathbb R}^n$ and $t\in [0,\infty)$ , these results also widen the range of weights in the known Riesz characterization of the classical weighted Hardy space $H^1_w({\mathbb R}^n)$ obtained by R. L. Wheeden from $w\in A_1({\mathbb R}^n)$ into $w\in A_\infty({\mathbb R}^n)$ with the sharp range $q(w)\in [1,\frac n{n-1})$ , where $q(w)$ denotes the critical index of the weight $w$ .

preprint2014arXiv

Weak Hardy Spaces $WH_L^p({\mathbb R}^n)$ Associated to Operators Satisfying $k$ -Davies-Gaffney Estimates

Let $L$ be a one-to-one operator of type $ω$ having a bounded $H_\infty$ functional calculus and satisfying the $k$ -Davies-Gaffney estimates with $k\in{\mathbb N}$ . In this paper, the authors introduce the weak Hardy space $WH_L^p(\mathbb{R}^n)$ associated to $L$ for $p\in (0,\,1]$ via the non-tangential square function $S_L$ and establish a weak molecular characterization of $WH_L^p(\mathbb{R}^n)$ . Typical examples of such operators include the $2k$ -order divergence form homogeneous elliptic operator $L_1:=(-1)^k\sum_{|α|=k=|β|}\partial^β(a_{α,β}\partial^α)$ , where $\{a_{α,β}\}_{|α|=k=|β|}$ are complex bounded measurable functions, and the $2k$ -order Schrödinger type operator $L_2:= (-Δ)^k+V^k$ , where $Δ$ is the Laplacian operator and $0\le V\in L^k_{\mathop\mathrm{loc}}(\mathbb{R}^n)$ . As applications, for $i\in\{1,\,2\}$ and $p\in(\frac{n}{n+k},\,1]$ , the authors prove that the associated Riesz transform $\nabla^k (L_i^{-1/2})$ is bounded from $WH^p_{L_i}(\mathbb{R}^n)$ to the classical weak Hardy space $WH^p(\mathbb{R}^n)$ and, for all $0<p<r\le1$ and $α=n(\frac{1}{p}-\frac{1}{r})$ , the fractional power $L_i^{-\fracα{2k}}$ is bounded from $WH_{L_i}^p(\mathbb{R}^n)$ to $WH_{L_i}^r(\mathbb{R}^n)$ . Furthermore, the authors find the dual space of $WH_L^p(\mathbb{R}^n)$ for $p\in(0,\,1]$ , which can be defined via mean oscillations based on some subtle coverings of bounded open sets and, even when $L:=-Δ$ , are also previously unknown. In particular, if $L$ is a nonnegative self-adjoint operator in $L^2({\mathbb R}^n)$ satisfying the Davies-Gaffney estimates, the authors further establish the weak atomic characterization of $WH_L^p(\mathbb{R}^n)$ .

preprint2012arXiv

Integrability conditions for the Grushin and Martinet distributions

We realise the first and second Grushin distributions as symmetry reductions of the 3-dimensional Heisenberg distribution and 4-dimensional Engel distribution respectively. Similarly, we realise the Martinet distribution as an alternative symmetry reduction of the Engel distribution. These reductions allow us to derive the integrability conditions for the Grushin and Martinet distributions and build certain complexes of differential operators. These complexes are well-behaved despite the distributions they resolve being non-regular.

preprint2012arXiv

Weighted Local Orlicz-Hardy Spaces on Domains and Their Applications in Inhomogeneous Dirichlet and Neumann Problems

Let $Ω$ be either $\mathbb{R}^n$ or a strongly Lipschitz domain of $\mathbb{R}^n$ , and $ω\in A_{\infty}(\mathbb{R}^n)$ (the class of Muckenhoupt weights). Let $L$ be a second order divergence form elliptic operator on $L^2 (Ω)$ with the Dirichlet or Neumann boundary condition, and assume that the heat semigroup generated by $L$ has the Gaussian property $(G_1)$ with the regularity of their kernels measured by $μ\in(0,1]$ . Let $Φ$ be a continuous, strictly increasing, subadditive, positive and concave function on $(0,\infty)$ of critical lower type index $p_Φ^-\in(0,1]$ . In this paper, the authors introduce the "geometrical" weighted local Orlicz-Hardy spaces $h^Φ_{ω,\,r}(Ω)$ and $h^Φ_{ω,\,z}(Ω)$ via the weighted local Orlicz-Hardy spaces $h^Φ_ω(\mathbb{R}^n)$ , and obtain their two equivalent characterizations in terms of the nontangential maximal function and the Lusin area function associated with the heat semigroup generated by $L$ when $p_Φ^-\in(n/(n+μ),1]$ . As applications, the authors prove that the operators $\nabla^2{\mathbb G}_D$ are bounded from $h^Φ_{ω,\,r}(Ω)$ to the weighted Orlicz space $L^Φ_ω(Ω)$ , and from $h^Φ_{ω,\,r}(Ω)$ to itself when $Ω$ is a bounded semiconvex domain in $\mathbb{R}^n$ and $p_Φ^-\in(\frac{n}{n+1},1]$ , and the operators $\nabla^2{\mathbb G}_N$ are bounded from $h^Φ_{ω,\,z}(Ω)$ to $L^Φ_ω(Ω)$ , and from $h^Φ_{ω,\,z}(Ω)$ to $h^Φ_{ω,\,r}(Ω)$ when $Ω$ is a bounded convex domain in $\mathbb{R}^n$ and $p_Φ^-\in(\frac{n}{n+1},1]$ , where ${\mathbb G}_D$ and ${\mathbb G}_N$ denote, respectively, the Dirichlet Green operator and the Neumann Green operator.

preprint2010arXiv

Hopf fibration: geodesics and distances

Here we study geodesics connecting two given points on odd-dimensional spheres respecting the Hopf fibration. This geodesic boundary value problem is completely solved in the case of 3-dimensional sphere and some partial results are obtained in the general case. The Carnot-Carathéodory distance is calculated. We also present some motivations related to quantum mechanics.

Der-Chen Chang

Research overview

Is this your work?

11 publications

The Analysis from Nonlinear Distance Metric to Kernel-based Drug Prescription Prediction System

Cross-Global Attention Graph Kernel Network Prediction of Drug Prescription

Gradient Estimates via Rearrangements for Solutions of Some Schrödinger Equations

Littlewood-Paley Characterizations of Hajłasz-Sobolev and Triebel-Lizorkin Spaces via Averages on Balls

On Li-Yau gradient estimate for sum of squares of vector fields up to higher step

On the Hodge-type decomposition and cohomolgy groups of $k$ -Cauchy-Fueter complexes over domains in the quaternionic space

Riesz Transform Characterizations of Musielak-Orlicz-Hardy Spaces

Weak Hardy Spaces $WH_L^p({\mathbb R}^n)$ Associated to Operators Satisfying $k$ -Davies-Gaffney Estimates

Integrability conditions for the Grushin and Martinet distributions

Weighted Local Orlicz-Hardy Spaces on Domains and Their Applications in Inhomogeneous Dirichlet and Neumann Problems

Hopf fibration: geodesics and distances

See the researcher in context

Building this map preview

Der-Chen Chang

Research overview

Is this your work?

11 publications

The Analysis from Nonlinear Distance Metric to Kernel-based Drug Prescription Prediction System

Cross-Global Attention Graph Kernel Network Prediction of Drug Prescription

Gradient Estimates via Rearrangements for Solutions of Some Schrödinger Equations

Littlewood-Paley Characterizations of Hajłasz-Sobolev and Triebel-Lizorkin Spaces via Averages on Balls

On Li-Yau gradient estimate for sum of squares of vector fields up to higher step

On the Hodge-type decomposition and cohomolgy groups of kkk-Cauchy-Fueter complexes over domains in the quaternionic space

Riesz Transform Characterizations of Musielak-Orlicz-Hardy Spaces

Weak Hardy Spaces WHLp(Rn)WH_L^p({\mathbb R}^n)WHLp​(Rn) Associated to Operators Satisfying kkk-Davies-Gaffney Estimates

Integrability conditions for the Grushin and Martinet distributions

Weighted Local Orlicz-Hardy Spaces on Domains and Their Applications in Inhomogeneous Dirichlet and Neumann Problems

Hopf fibration: geodesics and distances

See the researcher in context

Building this map preview

On the Hodge-type decomposition and cohomolgy groups of $k$ -Cauchy-Fueter complexes over domains in the quaternionic space

Weak Hardy Spaces $WH_L^p({\mathbb R}^n)$ Associated to Operators Satisfying $k$ -Davies-Gaffney Estimates