On the Limits of Sparsification
نویسندگان
چکیده
Impagliazzo, Paturi and Zane (JCSS 2001) proved a sparsification lemma for k-CNFs: every k-CNF is a sub-exponential size disjunction of k-CNFs with a linear number of clauses. This lemma has subsequently played a key role in the study of the exact complexity of the satisfiability problem. A natural question is whether an analogous structural result holds for CNFs or even for broader non-uniform classes such as constant-depth circuits or Boolean formulae. We prove a very strong negative result in this connection: For every superlinear function f(n), there are CNFs of size f(n) which cannot be written as a disjunction of 2n−εn CNFs each having a linear number of clauses for any ε > 0. We also give a hierarchy of such non-sparsifiable CNFs: For every k, there is a k′ for which there are CNFs of size n ′ which cannot be written as a sub-exponential size disjunction of CNFs of size n. Furthermore, our lower bounds hold not just against CNFs but against an arbitrary family of functions as long as the cardinality of the family is appropriately bounded. As by-products of our result, we make progress both on questions about circuit lower bounds for depth-3 circuits and satisfiability algorithms for constant-depth circuits. Improving on a result of Impagliazzo, Paturi and Zane, for any f(n) = ω(n log(n)), we define a pseudo-random function generator with seed length f(n) such that with high probability, a function in the output of this generator does not have depth-3 circuits of size 2n−o(n) with bounded bottom fan-in. We show that if we could decrease the seed length of our generator below n, we would get an explicit function which does not have linear-size logarithmic-depth series-parallel circuits, solving a long-standing open question. Motivated by the question of whether CNFs sparsify into boundeddepth circuits, we show a simplification result for bounded-depth circuits: any bounded-depth circuit of linear size can be written as a subexponential size disjunction of linear-size constant-width CNFs. As a corollary, we show that if there is an algorithm for CNF satisfiability which runs in time O(2) for some fixed α < 1 on CNFs of linear size, then there is an algorithm for satisfiability of linear-size constant-depth circuits which runs in time O(2). ? This is an extended abstract with some proofs missing. The full version may be found at [11]. ?? Partially supported by ESPRC Grant EP/H05068X/1 ? ? ? Work partially done as a Member at the Institute of Advanced Study, Princeton.
منابع مشابه
Performance Limits of Compressive Sensing Channel Estimation in Dense Cloud RAN
Towards reducing the training signaling overhead in large scale and dense cloud radio access networks (CRAN), various approaches have been proposed based on the channel sparsification assumption, namely, only a small subset of the deployed remote radio heads (RRHs) are of significance to any user in the system. Motivated by the potential of compressive sensing (CS) techniques in this setting, t...
متن کاملUnsupervised Sparsification of Similarity Graphs
Cluster analysis technology often grapples with high-dimensional and noisy data. The paper in hand identifies sparsification as an approach to address this problem. Sparsification improves both the runtime and the quality of cluster algorithms that exploit pairwise object similarities, i.e., that rely on similarity graphs. Sparsification has been addressed in the field of graphical cluster algo...
متن کاملSingle- and multi-level network sparsification by algebraic distance
Network sparsification methods play an important role in modern network analysis when fast estimation of computationally expensive properties (such as the diameter, centrality indices, and paths) is required. We propose a method of network sparsification that preserves a wide range of structural properties. Depending on the analysis goals, the method allows to distinguish between local and glob...
متن کاملDrawing Big Graphs Using Spectral Sparsification
Spectral sparsification is a general technique developed by Spielman et al. to reduce the number of edges in a graph while retaining its structural properties. We investigate the use of spectral sparsification to produce good visual representations of big graphs. We evaluate spectral sparsification approaches on real-world and synthetic graphs. We show that spectral sparsifiers are more effecti...
متن کاملSparse Approximations to Value Functions in Reinforcement Learning
We present a novel sparsification and value function approximation method for on-line reinforcement learning in continuous state and action spaces. Our approach is based on the kernel least squares temporal difference learning algorithm. We derive a recursive version and enhance the algorithm with a new sparsification mechanism based on the topology obtained from proximity graphs. The sparsific...
متن کاملMulti-resolutive sparse approximations of d-dimensional data
This paper proposes an iterative computation of sparse representations of functions defined on Rd , which exploits a formulation of the sparsification problem equivalent to Support Vector Machine and based on Tikhonov regularization. Through this equivalent formulation, the sparsification reduces to an approximation problem with a Tikhonov regularizer, which selects the null coefficients of the...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
- Electronic Colloquium on Computational Complexity (ECCC)
دوره 18 شماره
صفحات -
تاریخ انتشار 2011