arXiv subjects

Eric Zhang

Publications and source records attributed to Eric Zhang.

24 records · Page 2Linked to original sources

arXiv2023

A Stability Timescale for Non-Hierarchical Three-Body Systems

The gravitational three-body problem is a fundamental problem in physics and has significant applications to astronomy. Three-body configurations are often considered stable as long the system is hierarchical; that is, the two orbital distances are well-separated. However, instability, which is often associated with significant energy exchange between orbits, takes time to develop. Assuming two massive objects in a circular orbit and a test particle in an eccentric orbit, we develop an analytical formula estimating the time it takes for the test particle's orbital energy to change by an order of itself. We show its consistency with results from N-body simulations. For eccentric orbits in particular, the instability is primarily driven not by close encounters of the test particle with one of the other bodies, but by the fundamental susceptibility of eccentric orbits to exchange energy at their periapsis. Motivated by recent suggestions that the galactic center may host an intermediate-mass black hole (IMBH) as a companion to the massive black hole Sgr A*, we use our timescale to explore the parameter space that could harbor an IMBH for the lifetime of the S-cluster of stars surrounding Sgr A*. Furthermore, we show that the orbit of an S-star can be stable for long timescales in the presence of other orbital crossing stars, thus suggesting that the S-cluster may be stable for the lifetimes of its member stars.

astro-ph.GA↗

arXiv2023

Zero-shot Generative Model Adaptation via Image-specific Prompt Learning

Recently, CLIP-guided image synthesis has shown appealing performance on adapting a pre-trained source-domain generator to an unseen target domain. It does not require any target-domain samples but only the textual domain labels. The training is highly efficient, e.g., a few minutes. However, existing methods still have some limitations in the quality of generated images and may suffer from the mode collapse issue. A key reason is that a fixed adaptation direction is applied for all cross-domain image pairs, which leads to identical supervision signals. To address this issue, we propose an Image-specific Prompt Learning (IPL) method, which learns specific prompt vectors for each source-domain image. This produces a more precise adaptation direction for every cross-domain image pair, endowing the target-domain generator with greatly enhanced flexibility. Qualitative and quantitative evaluations on various domains demonstrate that IPL effectively improves the quality and diversity of synthesized images and alleviates the mode collapse. Moreover, IPL is independent of the structure of the generative model, such as generative adversarial networks or diffusion models. Code is available at https://github.com/Picsart-AI-Research/IPL-Zero-Shot-Generative-Model-Adaptation.

cs.CV↗

arXiv2023

Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models

The unlearning problem of deep learning models, once primarily an academic concern, has become a prevalent issue in the industry. The significant advances in text-to-image generation techniques have prompted global discussions on privacy, copyright, and safety, as numerous unauthorized personal IDs, content, artistic creations, and potentially harmful materials have been learned by these models and later utilized to generate and distribute uncontrolled content. To address this challenge, we propose \textbf{Forget-Me-Not}, an efficient and low-cost solution designed to safely remove specified IDs, objects, or styles from a well-configured text-to-image model in as little as 30 seconds, without impairing its ability to generate other content. Alongside our method, we introduce the \textbf{Memorization Score (M-Score)} and \textbf{ConceptBench} to measure the models' capacity to generate general concepts, grouped into three primary categories: ID, object, and style. Using M-Score and ConceptBench, we demonstrate that Forget-Me-Not can effectively eliminate targeted concepts while maintaining the model's performance on other concepts. Furthermore, Forget-Me-Not offers two practical extensions: a) removal of potentially harmful or NSFW content, and b) enhancement of model accuracy, inclusion and diversity through \textbf{concept correction and disentanglement}. It can also be adapted as a lightweight model patch for Stable Diffusion, allowing for concept manipulation and convenient distribution. To encourage future research in this critical area and promote the development of safe and inclusive generative models, we will open-source our code and ConceptBench at \href{https://github.com/SHI-Labs/Forget-Me-Not}{https://github.com/SHI-Labs/Forget-Me-Not}.

cs.CV↗

arXiv2021

Limit Densities of Patterns in Permutation Inflations

Call a permutation $k$-inflatable if the sequence of its tensor products with uniform random permutations of increasing lengths has uniform $k$-point pattern densities. Previous work has shown that nontrivial $k$-inflatable permutations do not exist for $k \geq 4$. In this paper, we derive a general formula for the limit densities of patterns in the sequence of tensor products of a fixed permutation with each permutation from a convergent sequence. By applying this result, we completely characterize $3$-inflatable permutations and find explicit examples of $3$-inflatable permutations with various lengths, including the shortest examples with length $17$.

math.CO↗

arXiv2018

TrueChain: Highly Performant Decentralized Public Ledger

In this paper we present the initial design of Minerva consensus protocol for Truechain and other technical details. Currently, it is widely believed in the blockchain community that a public chain cannot simultaneously achieve high performance, decentralization and security. This is true in the case of a Nakamoto chain (low performance) or a delegated proof of stake chain (partially centralized), which are the most popular block chain solutions at time of writing. Our consensus design enjoys the same consistency, liveness, transaction finality and security guarantee, a de-facto with the Hybrid Consensus. We go on to propose the idea of a new virtual machine on top of Ethereum which adds permissioned-chain based transaction processing capabilities in a permissionless setting. We also use the idea of data sharding and speculative transactions, and evaluation of smart contracts in a sharding friendly virtual machine. Finally, we will briefly discuss our fundamentally ASIC resistant mining algorithm, Truehash.

cs.DC↗

arXiv2014

Impact of Energy Consumption on Multipath TCP Enabled Mobiles

Multiple accesses are common for most mobile devices today. This technological advance opens up a new design space for improving the communication performance of mobile devices. Multipath TCP is a TCP extension that enables using multiple network paths between two end systems for a single TCP connection, increasing performance and reliability. Meanwhile, when operating multiple active interfaces, multipath-TCP also consumes substantial more power and drains out bettery faster than using one interface. Thus, enabling Multipath TCP on mobile devices brings in new challenges. In this paper, we theoretically analyze the underlying design choices given by the Multipath TCP. In particular, we theoretically formulate the rela- tion between performance (throughput) and energy consumption for Multipath TCP. We find that sometime the throughput and energy consumption can be concurrently improved.

cs.NI↗