Searcharxiv⌕ Search

arXiv subjects

Haonan Zhang

Publications and source records attributed to Haonan Zhang.

At least 73 records · Page 4Linked to original sources
arXiv2024

Text-Video Retrieval with Global-Local Semantic Consistent Learning

Adapting large-scale image-text pre-training models, e.g., CLIP, to the video domain represents the current state-of-the-art for text-video retrieval. The primary approaches involve transferring text-video pairs to a common embedding space and leveraging cross-modal interactions on specific entities for semantic alignment. Though effective, these paradigms entail prohibitive computational costs, leading to inefficient retrieval. To address this, we propose a simple yet effective method, Global-Local Semantic Consistent Learning (GLSCL), which capitalizes on latent shared semantics across modalities for text-video retrieval. Specifically, we introduce a parameter-free global interaction module to explore coarse-grained alignment. Then, we devise a shared local interaction module that employs several learnable queries to capture latent semantic concepts for learning fine-grained alignment. Furthermore, an Inter-Consistency Loss (ICL) is devised to accomplish the concept alignment between the visual query and corresponding textual query, and an Intra-Diversity Loss (IDL) is developed to repulse the distribution within visual (textual) queries to generate more discriminative concepts. Extensive experiments on five widely used benchmarks (i.e., MSR-VTT, MSVD, DiDeMo, LSMDC, and ActivityNet) substantiate the superior effectiveness and efficiency of the proposed method. Remarkably, our method achieves comparable performance with SOTA as well as being nearly 220 times faster in terms of computational cost. Code is available at: https://github.com/zchoi/GLSCL.

cs.CV↗
arXiv2024

Bohnenblust--Hille inequality for cyclic groups

For any $K>2$ and the multiplicative cyclic group $Ω_K$ of order $K$, consider any function $f:Ω_K^n\to\mathbf{C}$ and its Fourier expansion $f(z)=\sum_{α\in\{0,1,\ldots,K-1\}^n}a_αz^α$, with $d:=\text{deg}(f)$ denoting its degree as a multivariate polynomial. We prove a Bohnenblust--Hille (BH) inequality in this setting: the $\ell_{2d/(d+1)}$ norm of the Fourier coefficients of $f$ is bounded by $C(d,K)\|f\|_\infty$ with $C(d,K)$ independent of $n$. This is the interpolating case between the now well-understood BH inequalities for functions on the poly-torus ($K =\infty$) and the hypercube ($K=2$) but those extreme cases of $K$ have special properties whose absence for intermediate $K$ prevent a proof by the standard BH framework. New techniques are developed exploiting the group structure of $Ω_K^n$. By known reductions, the cyclic group BH inequality also entails a noncommutative BH inequality for tensor products of the $K \times K$ complex matrix algebra (or in the language of quantum mechanics, systems of $K$-level qudits). These new BH inequalities generalize several applications in harmonic analysis and statistical learning theory to broader classes of functions and operators.

math.FA↗
arXiv2024

Noncommutative Bohnenblust--Hille Inequality for qudit systems

Previous noncommutative Bohnenblust--Hille (BH) inequalities addressed operator decompositions in the tensor-product space $M_2(\mathbb{C})^{\otimes n}$; \emph{i.e.,} for systems of qubits \cite{HCP22,VZ23}. Here we prove noncommutative BH inequalities for operators decomposed in tensor-product spaces of arbitrary local dimension, \emph{i.e.,} $M_K(\mathbb{C})^{\otimes n}$ for any $K\geq2$ or on systems of $K$-level qudits. We treat operator decompositions in both the Gell-Mann and Heisenberg--Weyl basis, reducing to the recently-proved commutative hypercube BH \cite{DMP} and cyclic group BH \cite{SVZ} inequalities respectively. As an application we discuss learning qudit quantum observables.

math.FA↗
arXiv2024

Quantum Talagrand, KKL and Friedgut's theorems and the learnability of quantum Boolean functions

We extend three related results from the analysis of influences of Boolean functions to the quantum setting, namely the KKL Theorem, Friedgut's Junta Theorem and Talagrand's variance inequality for geometric influences. Our results are derived by a joint use of recently studied hypercontractivity and gradient estimates. These generic tools also allow us to derive generalizations of these results in a general von Neumann algebraic setting beyond the case of the quantum hypercube, including examples in infinite dimensions relevant to quantum information theory such as continuous variables quantum systems. Finally, we comment on the implications of our results as regards to noncommutative extensions of isoperimetric type inequalities, quantum circuit complexity lower bounds and the learnability of quantum observables.

math.FA↗
arXiv2024

Intertwining Curvature Bounds for Graphs and Quantum Markov Semigroups

Based on earlier work by Carlen-Maas and the second- and third-named author, we introduce the notion of intertwining curvature lower bounds for graphs and quantum Markov semigroups. This curvature notion is stronger than both Bakry-Émery and entropic Ricci curvature, while also computationally simpler than the latter. We verify intertwining curvature bounds in a number of examples, including finite weighted graphs and graphs with Laplacians admitting nice mapping representations, as well as generalized dephasing semigroups and quantum Markov semigroups whose generators are formed by commuting jump operators. By improving on the best-known bounds for entropic curvature of depolarizing semigroups, we demonstrate that there can be a gap between the optimal intertwining and entropic curvature bound. In the case of qubits, this improved entropic curvature bound implies the modified logarithmic Sobolev inequality with optimal constant.

math.FA↗
arXiv2023

IS-DARTS: Stabilizing DARTS through Precise Measurement on Candidate Importance

Among existing Neural Architecture Search methods, DARTS is known for its efficiency and simplicity. This approach applies continuous relaxation of network representation to construct a weight-sharing supernet and enables the identification of excellent subnets in just a few GPU days. However, performance collapse in DARTS results in deteriorating architectures filled with parameter-free operations and remains a great challenge to the robustness. To resolve this problem, we reveal that the fundamental reason is the biased estimation of the candidate importance in the search space through theoretical and experimental analysis, and more precisely select operations via information-based measurements. Furthermore, we demonstrate that the excessive concern over the supernet and inefficient utilization of data in bi-level optimization also account for suboptimal results. We adopt a more realistic objective focusing on the performance of subnets and simplify it with the help of the information-based measurements. Finally, we explain theoretically why progressively shrinking the width of the supernet is necessary and reduce the approximation error of optimal weights in DARTS. Our proposed method, named IS-DARTS, comprehensively improves DARTS and resolves the aforementioned problems. Extensive experiments on NAS-Bench-201 and DARTS-based search space demonstrate the effectiveness of IS-DARTS.

cs.LG↗
arXiv2023

Quantum and classical low-degree learning via a dimension-free Remez inequality

Recent efforts in Analysis of Boolean Functions aim to extend core results to new spaces, including to the slice $\binom{[n]}{k}$, the hypergrid $[K]^n$, and noncommutative spaces (matrix algebras). We present here a new way to relate functions on the hypergrid (or products of cyclic groups) to their harmonic extensions over the polytorus. We show the supremum of a function $f$ over products of the cyclic group $\{\exp(2πi k/K)\}_{k=1}^K$ controls the supremum of $f$ over the entire polytorus $(\{z\in\mathbf{C}:|z|=1\}^n)$, with multiplicative constant $C$ depending on $K$ and $\text{deg}(f)$ only. This Remez-type inequality appears to be the first such estimate that is dimension-free (i.e., $C$ does not depend on $n$). This dimension-free Remez-type inequality removes the main technical barrier to giving $\mathcal{O}(\log n)$ sample complexity, polytime algorithms for learning low-degree polynomials on the hypergrid and low-degree observables on level-$K$ qudit systems. In particular, our dimension-free Remez inequality implies new Bohnenblust--Hille-type estimates which are central to the learning algorithms and appear unobtainable via standard techniques. Thus we extend to new spaces a recent line of work \cite{EI22, CHP, VZ22} that gave similarly efficient methods for learning low-degree polynomials on the hypercube and observables on qubits. An additional product of these efforts is a new class of distributions over which arbitrary quantum observables are well-approximated by their low-degree truncations -- a phenomenon that greatly extends the reach of low-degree learning in quantum science \cite{CHP}.

math.AP↗
arXiv2023

RAH! RecSys-Assistant-Human: A Human-Centered Recommendation Framework with LLM Agents

The rapid evolution of the web has led to an exponential growth in content. Recommender systems play a crucial role in Human-Computer Interaction (HCI) by tailoring content based on individual preferences. Despite their importance, challenges persist in balancing recommendation accuracy with user satisfaction, addressing biases while preserving user privacy, and solving cold-start problems in cross-domain situations. This research argues that addressing these issues is not solely the recommender systems' responsibility, and a human-centered approach is vital. We introduce the RAH Recommender system, Assistant, and Human) framework, an innovative solution with LLM-based agents such as Perceive, Learn, Act, Critic, and Reflect, emphasizing the alignment with user personalities. The framework utilizes the Learn-Act-Critic loop and a reflection mechanism for improving user alignment. Using the real-world data, our experiments demonstrate the RAH framework's efficacy in various recommendation domains, from reducing human burden to mitigating biases and enhancing user control. Notably, our contributions provide a human-centered recommendation framework that partners effectively with various recommendation models.

cs.IR↗
arXiv2023

Noncommutative Bohnenblust--Hille inequalities

Bohnenblust--Hille inequalities for Boolean cubes have been proven with dimension-free constants that grow subexponentially in the degree \cite{defant2019fourier}. Such inequalities have found great applications in learning low-degree Boolean functions \cite{eskenazis2022learning}. Motivated by learning quantum observables, a qubit analogue of Bohnenblust--Hille inequality for Boolean cubes was recently conjectured in \cite{RWZ22}. The conjecture was resolved in \cite{CHP}. In this paper, we give a new proof of these Bohnenblust--Hille inequalities for qubit system with constants that are dimension-free and of exponential growth in the degree. As a consequence, we obtain a junta theorem for low-degree polynomials. Using similar ideas, we also study learning problems of low degree quantum observables and Bohr's radius phenomenon on quantum Boolean cubes.

math.FA↗
arXiv2023

Some convexity and monotonicity results of trace functionals

In this paper, we prove the convexity of trace functionals $$(A,B,C)\mapsto \text{Tr}|B^{p}AC^{q}|^{s},$$ for parameters $(p,q,s)$ that are best possible, where $B$ and $C$ are any $n$-by-$n$ positive definite matrices, and $A$ is any $n$-by-$n$ matrix. We also obtain the monotonicity versions of trace functionals of this type. As applications, we extend some results in \cite{HP12quasi,CFL16some} and resolve a conjecture in \cite{RZ14} in the matrix setting. Other conjectures in \cite{RZ14} will also be discussed. We also show that some related trace functionals are not concave in general. Such concavity results were expected to hold in different problems.

math-ph↗
arXiv2023

Self-supervised arbitrary scale super-resolution framework for anisotropic MRI

In this paper, we propose an efficient self-supervised arbitrary-scale super-resolution (SR) framework to reconstruct isotropic magnetic resonance (MR) images from anisotropic MRI inputs without involving external training data. The proposed framework builds a training dataset using in-the-wild anisotropic MR volumes with arbitrary image resolution. We then formulate the 3D volume SR task as a SR problem for 2D image slices. The anisotropic volume's high-resolution (HR) plane is used to build the HR-LR image pairs for model training. We further adapt the implicit neural representation (INR) network to implement the 2D arbitrary-scale image SR model. Finally, we leverage the well-trained proposed model to up-sample the 2D LR plane extracted from the anisotropic MR volumes to their HR views. The isotropic MR volumes thus can be reconstructed by stacking and averaging the generated HR slices. Our proposed framework has two major advantages: (1) It only involves the arbitrary-resolution anisotropic MR volumes, which greatly improves the model practicality in real MR imaging scenarios (e.g., clinical brain image acquisition); (2) The INR-based SR model enables arbitrary-scale image SR from the arbitrary-resolution input image, which significantly improves model training efficiency. We perform experiments on a simulated public adult brain dataset and a real collected 7T brain dataset. The results indicate that our current framework greatly outperforms two well-known self-supervised models for anisotropic MR image SR tasks.

eess.IV↗
arXiv2023

Leveraging policy instruments and financial incentives to reduce embodied carbon in energy retrofits

The existing buildings and building construction sectors together are responsible for over one-third of the total global energy consumption and nearly 40% of total greenhouse gas (GHG) emissions. GHG emissions from the building sector are made up of embodied emissions and operational emissions. Recognizing the importance of reducing energy use and emissions associated with the building sector, governments have introduced policies, standards, and design guidelines to improve building energy performance and reduce GHG emissions associated with operating buildings. However, policy initiatives that reduce embodied emissions of the existing building sector are lacking. This research aims to develop policy strategies to reduce embodied carbon emissions in retrofits. In order to achieve this goal, this research conducted a literature review and identification of policies and financial incentives in British Columbia (BC) for reducing overall GHG emissions from the existing building sector. Then, this research analyzed worldwide policies and incentives that reduce embodied carbon emissions in the existing building sector. After reviewing the two categories of retrofit policies, the author identified links and opportunities between existing BC strategies, tools, and incentives, and global embodied emission strategies. Finally, this research compiled key findings from all resources and provided policy recommendations for reducing embodied carbon emissions in retrofits in BC.

econ.GN↗
arXiv2023

Life cycle costing analysis of deep energy retrofits of a mid-rise building to understand the impact of energy conservation measures

Building energy retrofits have been identified as key to realizing climate mitigation goals in Canada. This study aims to provide a roadmap for existing mid-rise building retrofits in order to understand the required capital investment, energy savings, energy cost savings, and carbon footprint for mid-rise residential buildings in Canada. This study employed EnergyPlus to examine the energy performance of 11 energy retrofit measures for a typical multi-unit residential building (MURB) in Metro Vancouver, British Columbia, Canada. The author employed the energy simulation software (EnergyPlus) to evaluate the pre-and post-retrofit operational energy performance of the selected MURB. Two base building models powered by natural gas (NG-building) and electricity (E-building) were created by SketchUP. The energy simulation results were combined with cost and emission impact data to evaluate the economic and environmental performance of the selected energy retrofit measures. The results indicated that the NG-building can produce significant GHG emission reductions (from 27.64 tCO2e to 3.77 tCO2e) by implementing these energy retrofit measures. In terms of energy savings, solar PV, ASHP, water heater HP, and HRV enhancement have great energy saving potential compared to other energy retrofit measures. In addition, temperature setback, lighting, and airtightness enhancement present the best economic performance from a life cycle perspective. However, windows, ASHP, and solar PV, are not economical choices because of higher life cycle costs. While ASHP can increase life cycle costs for the NG-building, with the financial incentives provided by the governments, ASHP could be the best choice to reduce GHG emissions when stakeholders make decisions on implementing energy retrofits.

econ.GN↗
arXiv2023

A Finite Blocklength Approach for Wireless Hierarchical Federated Learning in the Presence of Physical Layer Security

In this paper, the wireless hierarchical federated learning (HFL) is revisited by considering physical layer security (PLS). First, we establish a framework for this new problem. Then, we propose a practical finite blocklength (FBL) coding scheme for the wireless HFL in the presence of PLS, which is self-secure when the coding blocklength is lager than a certain threshold. Finally, the study of this paper is further explained via numerical examples and simulation results.

cs.IT↗
arXiv2022

Visual Commonsense-aware Representation Network for Video Captioning

Generating consecutive descriptions for videos, i.e., Video Captioning, requires taking full advantage of visual representation along with the generation process. Existing video captioning methods focus on making an exploration of spatial-temporal representations and their relationships to produce inferences. However, such methods only exploit the superficial association contained in the video itself without considering the intrinsic visual commonsense knowledge that existed in a video dataset, which may hinder their capabilities of knowledge cognitive to reason accurate descriptions. To address this problem, we propose a simple yet effective method, called Visual Commonsense-aware Representation Network (VCRN), for video captioning. Specifically, we construct a Video Dictionary, a plug-and-play component, obtained by clustering all video features from the total dataset into multiple clustered centers without additional annotation. Each center implicitly represents a visual commonsense concept in the video domain, which is utilized in our proposed Visual Concept Selection (VCS) to obtain a video-related concept feature. Next, a Conceptual Integration Generation (CIG) is proposed to enhance the caption generation. Extensive experiments on three publicly video captioning benchmarks: MSVD, MSR-VTT, and VATEX, demonstrate that our method reaches state-of-the-art performance, indicating the effectiveness of our method. In addition, our approach is integrated into the existing method of video question answering and improves this performance, further showing the generalization of our method. Source code has been released at https://github.com/zchoi/VCRN.

cs.CV↗
arXiv2022

Heat-smoothing for holomorphic subalgebras of free group von Neumann algebras

The heat semigroup on discrete hypercubes is well-known to be contractive over $L_p$-spaces for $1<p<\infty$. A question of Mendel and Naor \cite{MN14} concerns a stronger contraction property in the tail spaces, which is known as the heat-smoothing conjecture. Eskenazis and Ivanisvili \cite{EI20} considered a Gaussian analog of this conjecture and resolved some special cases. In particular, they proved that heat-smoothing type conjecture holds for holomorphic functions in the Gaussian spaces with sharp constants. In this paper, we prove analogous sharp inequalities for holomorphic subalgebras of free group von Neumann algebras. Similar results also hold for $q$-Gaussian algebras and quantum tori. In the case of free group von Neumann algebras, the weaker formulation of heat-smoothing is proved with optimal order.

math.OA↗
arXiv2022

Monotonicity versions of Epstein's Concavity Theorem and related inequalities

Many trace inequalities can be expressed either as concavity/convexity theorems or as monotonicity theorems. A classic example is the joint convexity of the quantum relative entropy which is equivalent to the Data Processing Inequality. The latter says that quantum operations can never increase the relative entropy. The monotonicity versions often have many advantages, and often have direct physical application, as in the example just mentioned. Moreover, the monotonicity results are often valid for a larger class of maps than, say, quantum operations (which are completely positive). In this paper we prove several new monotonicity results, the first of which is a monotonicity theorem that has as a simple corollary a celebrated concavity theorem of Epstein. Our starting points are the monotonicity versions of the Lieb Concavity and the Lieb Convexity Theorems. We also give two new proofs of these in their general forms using interpolation. We then prove our new monotonicity theorems by several duality arguments.

math.FA↗