Searcharxiv⌕ Search

arXiv subjects

Mingmou Liu

Publications and source records attributed to Mingmou Liu.

9 recordsLinked to original sources
arXiv2026

SDDBMs: Soft Denoising Diffusion Bridge Models

Diffusion bridge models leverage Doob's \(h\)-transform to construct stochastic transports between arbitrary endpoint distributions, and have shown strong potential in image-to-image translation and restoration. However, most existing bridge models rely on hard endpoint conditioning, which forces the terminal state to match a prescribed target exactly. This hard constraint induces terminal-boundary singularities: the terminal law collapses to a Dirac measure, and the resulting drift coefficients become ill-conditioned near the endpoint. In this paper, we propose Soft Denoising Diffusion Bridge Models (SDDBMs), a generalized framework that regularizes diffusion bridges directly at the level of their terminal constraints. Instead of imposing an exact endpoint, SDDBMs prescribe a non-degenerate Gaussian terminal marginal under the transformed path measure, with a flexible terminal center and variance. Starting from this prescribed marginal, we develop a complete closed-form construction of the soft bridge, including the Gaussian terminal reweighting and soft \(h\)-function, the induced Gaussian forward marginals and \(\mathbf{x}_0\)-free dynamics. Theoretically, SDDBMs provide a unified probabilistic perspective that encompasses existing diffusion bridge models, including DDBMs, GOUB, and UniDB, as special cases under specific parameter choices. Extensive experiments on image restoration tasks demonstrate that SDDBMs achieve improved numerical stability and superior generation quality over existing bridge-based methods.

cs.AI↗
arXiv2023

Cell-Probe Lower Bound for Accessible Interval Graphs

We spot a hole in the area of succinct data structures for graph classes from a universe of size at most $n^n$. Very often, the input graph is labeled by the user in an arbitrary and easy-to-use way, and the data structure for the graph relabels the input graph in some way. For any access, the user needs to store these labels or compute the new labels in an online manner. This might require more bits than the information-theoretic minimum of the original graph class, hence, defeating the purpose of succinctness. Given this, the data structure designer must allow the user to access the data structure with the original labels, i.e., relabeling is not allowed. We call such a graph data structure ``accessible''. In this paper, we study the complexity of such accessible data structures for interval graphs, a graph class with information-theoretic minimum less than $n\log n$ bits. - We formalize the concept of "accessibility" (which was implicitly assumed), and propose the "universal interval representation", for interval graphs. - Any data structure for interval graphs in universal interval representation, which supports both adjacency and degree query simultaneously with time cost $t_1$ and $t_2$ respectively, must consume at least $\log_2(n!)+n/(\log n)^{O(t_1+t_2)}$ bits of space. This is also the first lower bound for graph classes with information-theoretic minimum less than $n\log_2n$ bits. - We provide efficient succinct data structures for interval graphs in universal interval representation supporting adjacency query and degree query individually in constant time and space costs. Therefore, two upper bounds together with the lower bound show that the two elementary queries for interval graphs are incompatible with each other in the context of succinct data structure. To the best of our knowledge, this is the first proof of such incompatibility phenomenon.

cs.DS↗
arXiv2023

Sparsity-Dimension Trade-Offs for Oblivious Subspace Embeddings

An oblivious subspace embedding (OSE), characterized by parameters $m,n,d,ε,δ$, is a random matrix $Π\in \mathbb{R}^{m\times n}$ such that for any $d$-dimensional subspace $T\subseteq \mathbb{R}^n$, $\Pr_Π[\forall x\in T, (1-ε)\|x\|_2 \leq \|Πx\|_2\leq (1+ε)\|x\|_2] \geq 1-δ$. When an OSE has $s\le 1/2.001ε$ nonzero entries in each column, we show it must hold that $m = Ω\left(d^2/( ε^2s^{1+O(δ)})\right)$, which is the first lower bound with multiplicative factors of $d^2$ and $1/ε$, improving on the previous $Ω\left(d^2/s^{O(δ)}\right)$ lower bound due to Li and Liu (PODS 2022). When an OSE has $s=Ω(\log(1/ε)/ε)$ nonzero entries in each column, we show it must hold that $m = Ω\left((d/ε)^{1+1/4.001εs}/s^{O(δ)}\right)$, which is the first lower bound with multiplicative factors of $d$ and $1/ε$, improving on the previous $Ω\left(d^{1+1/(16εs+4)}\right)$ lower bound due to Nelson and Nguyen (ICALP 2014). This second result is a special case of a more general trade-off among $d,ε,s,δ$ and $m$.

cs.DS↗
arXiv2022

Nearly Tight Lower Bounds for Succinct Range Minimum Query

Given an array of distinct integers $A[1\ldots n]$, the Range Minimum Query (RMQ) problem requires us to construct a data structure from $A$, supporting the RMQ query: given an interval $[a,b]\subseteq[1,n]$, return the index of the minimum element in subarray $A[a\ldots b]$, i.e. return $\text{argmin}_{i\in[a,b]}A[i]$. The fundamental problem has a long history. The textbook solution which uses $O(n)$ words of space and $O(1)$ time by Gabow, Bentley, Tarjan (STOC 1984) and Harel, Tarjan (SICOMP 1984) dates back to 1980s. The state-of-the-art solution is presented by Fischer, Heun (SICOMP 2011) and Navarro, Sadakane (TALG 2014). The solution uses $2n-1.5\log n+n/\left(\frac{\log n}{t}\right)^t+\tilde{O}(n^{3/4})$ bits of space and $O(t)$ query time, where the additive $\tilde{O}(n^{3/4})$ is a pre-computed lookup table used in the RAM model, assuming the word-size is $Θ(\log n)$ bits. On the other hand, the only known lower bound is proved by Liu and Yu (STOC 2020). They show that any data structure which solves RMQ in $t$ query time must use $2n-1.5\log n+n/(\log n)^{O(t^2\log^2t)}$ bits of space, assuming the word-size is $Θ(\log n)$ bits. In this paper, we prove nearly tight lower bound for this problem. We show that, for any data structure which solves RMQ in $t$ query time, $2n-1.5\log n+n/(\log n)^{O(t\log^2t)}$ bits of space is necessary in the cell-probe model with word-size $Θ(\log n)$ bits. We emphasize that, in terms of time complexity, our lower bound is tight up to a polylogarithmic factor.

cs.DS↗
arXiv2021

Lower Bounds for Sparse Oblivious Subspace Embeddings

An oblivious subspace embedding (OSE), characterized by parameters $m,n,d,ε,δ$, is a random matrix $Π\in \mathbb{R}^{m\times n}$ such that for any $d$-dimensional subspace $T\subseteq \mathbb{R}^n$, $\Pr_Π[\forall x\in T, (1-ε)\|x\|_2 \leq \|Πx\|_2\leq (1+ε)\|x\|_2] \geq 1-δ$. For $ε$ and $δ$ at most a small constant, we show that any OSE with one nonzero entry in each column must satisfy that $m = Ω(d^2/(ε^2δ))$, establishing the optimality of the classical Count-Sketch matrix. When an OSE has $1/(9ε)$ nonzero entries in each column, we show it must hold that $m = Ω(ε^{O(δ)} d^2)$, improving on the previous $Ω(ε^2 d^2)$ lower bound due to Nelson and Nguyen (ICALP 2014).

cs.DS↗
arXiv2021

On the Optimal Time/Space Tradeoff for Hash Tables

For nearly six decades, the central open question in the study of hash tables has been to determine the optimal achievable tradeoff curve between time and space. State-of-the-art hash tables offer the following guarantee: If keys/values are Theta(log n) bits each, then it is possible to achieve constant-time insertions/deletions/queries while wasting only O(loglog n) bits of space per key when compared to the information-theoretic optimum. Even prior to this bound being achieved, the target of O(loglog n) wasted bits per key was known to be a natural end goal, and was proven to be optimal for a number of closely related problems (e.g., stable hashing, dynamic retrieval, and dynamically-resized filters). This paper shows that O(loglog n) wasted bits per key is not the end of the line for hashing. In fact, for any k \in [log* n], it is possible to achieve O(k)-time insertions/deletions, O(1)-time queries, and O(\log^{(k)} n) wasted bits per key (all with high probability in n). This means that, each time we increase insertion/deletion time by an \emph{additive constant}, we reduce the wasted bits per key \emph{exponentially}. We further show that this tradeoff curve is the best achievable by any of a large class of hash tables, including any hash table designed using the current framework for making constant-time hash tables succinct.

cs.DS↗
arXiv2020

Succinct Filters for Sets of Unknown Sizes

The membership problem asks to maintain a set $S\subseteq[u]$, supporting insertions and membership queries, i.e., testing if a given element is in the set. A data structure that computes exact answers is called a dictionary. When a (small) false positive rate $ε$ is allowed, the data structure is called a filter. The space usages of the standard dictionaries or filters usually depend on the upper bound on the size of $S$, while the actual set can be much smaller. Pagh, Segev and Wieder (FOCS'13) were the first to study filters with varying space usage based on the current $|S|$. They showed in order to match the space with the current set size $n=|S|$, any filter data structure must use $(1-o(1))n(\log(1/ε)+(1-O(ε))\log\log n)$ bits, in contrast to the well-known lower bound of $N\log(1/ε)$ bits, where $N$ is an upper bound on $|S|$. They also presented a data structure with almost optimal space of $(1+o(1))n(\log(1/ε)+O(\log\log n))$ bits provided that $n>u^{0.001}$, with expected amortized constant insertion time and worst-case constant lookup time. In this work, we present a filter data structure with improvements in two aspects: - it has constant worst-case time for all insertions and lookups with high probability; - it uses space $(1+o(1))n(\log (1/ε)+\log\log n)$ bits when $n>u^{0.001}$, achieving optimal leading constant for all $ε=o(1)$. We also present a dictionary that uses $(1+o(1))n\log(u/n)$ bits of space, matching the optimal space in terms of the current size, and performs all operations in constant time with high probability.

cs.DS↗
arXiv2020

Lower Bound for Succinct Range Minimum Query

Given an integer array $A[1..n]$, the Range Minimum Query problem (RMQ) asks to preprocess $A$ into a data structure, supporting RMQ queries: given $a,b\in [1,n]$, return the index $i\in[a,b]$ that minimizes $A[i]$, i.e., $\mathrm{argmin}_{i\in[a,b]} A[i]$. This problem has a classic solution using $O(n)$ space and $O(1)$ query time by Gabow, Bentley, Tarjan (STOC, 1984) and Harel, Tarjan (SICOMP, 1984). The best known data structure by Fischer, Heun (SICOMP, 2011) and Navarro, Sadakane (TALG, 2014) uses $2n+n/(\frac{\log n}{t})^t+\tilde{O}(n^{3/4})$ bits and answers queries in $O(t)$ time, assuming the word-size is $w=Θ(\log n)$. In particular, it uses $2n+n/\mathrm{poly}\log n$ bits of space as long as the query time is a constant. In this paper, we prove the first lower bound for this problem, showing that $2n+n/\mathrm{poly}\log n$ space is necessary for constant query time. In general, we show that if the data structure has query time $O(t)$, then it must use at least $2n+n/(\log n)^{\tilde{O}(t^2)}$ space, in the cell-probe model with word-size $w=Θ(\log n)$.

cs.DS↗
arXiv2016

Randomized approximate nearest neighbor search with limited adaptivity

We study the fundamental problem of approximate nearest neighbor search in $d$-dimensional Hamming space $\{0,1\}^d$. We study the complexity of the problem in the famous cell-probe model, a classic model for data structures. We consider algorithms in the cell-probe model with limited adaptivity, where the algorithm makes $k$ rounds of parallel accesses to the data structure for a given $k$. For any $k\ge 1$, we give a simple randomized algorithm solving the approximate nearest neighbor search using $k$ rounds of parallel memory accesses, with $O(k(\log d)^{1/k})$ accesses in total. We also give a more sophisticated randomized algorithm using $O(k+(\frac{1}{k}\log d)^{O(1/k)})$ memory accesses in $k$ rounds for large enough $k$. Both algorithms use data structures of size polynomial in $n$, the number of points in the database. For the lower bound, we prove an $Ω(\frac{1}{k}(\log d)^{1/k})$ lower bound for the total number of memory accesses required by any randomized algorithm solving the approximate nearest neighbor search within $k\le\frac{\log\log d}{2\log\log\log d}$ rounds of parallel memory accesses on any data structures of polynomial size. This lower bound shows that our first algorithm is asymptotically optimal for any constant round $k$. And our second algorithm approaches the asymptotically optimal tradeoff between rounds and memory accesses, in a sense that the lower bound of memory accesses for any $k_1$ rounds can be matched by the algorithm within $k_2=O(k_1)$ rounds. In the extreme, for some large enough $k=Θ\left(\frac{\log\log d}{\log\log\log d}\right)$, our second algorithm matches the $Θ\left(\frac{\log\log d}{\log\log\log d}\right)$ tight bound for fully adaptive algorithms for approximate nearest neighbor search due to Chakrabarti and Regev.

cs.DS↗