SearcharxivSearch

arXiv subjects

Chris Pollett

Publications and source records attributed to Chris Pollett.

5 recordsLinked to original sources

Arithmetics within the Linear Time Hierarchy

We identify fragments of the arithmetic $S_1$ that enjoy nice closure properties and have exact characterization of their definable multifunctions. To do this, in the language of $S_1$, $L_1$, starting from the formula classes, $Σ^{\mathsf b}_{i}$, which ignore sharply bounded quantifiers when determining quantifier alternations, we define new syntactic classes by counting bounded existential sharply bounded universal quantifiers blocks. Using these, we define arithmetics: $\breve{S}^{i}_{1}$, $TLS^i_1$ and $TSC^i_1$. $\breve{S}^{i}_{1}$ consists of open axioms for the language symbols and length induction for one of our new classes, $SIUT_{i,1}^{\{p(|id|)\}}$. $TLS^i_1$ and $TSC^i_1$ are defined using axioms related to dependent choice sequences for formulas from two other classes within $Σ^{\mathsf b}_{i}$. We prove for $i \geq 1$ that $$TLS^i_1 \subseteq TSC^i_1 \subseteq \breve{S}^{i}_{1} \preceq_{\forall B(SITT_{i+1}^{\{p(|id|)\}})} TLS^{i+1}_1$$ and that the $SITT_{i}^{\{p(|id|)\}}$-definable in $TLS^i_1$ (resp. $SITT_{i}^{\{2^{p(||id||)}\}}$-definable in $TSC^i_1$) multifunctions are $L_1$-$FLOGSPACE^{SIT_{i,1}}[wit]$ (resp. $L_1$-$FSC^{SIT_{i,1}}[wit]$). These multifunction classes are respectively the logspace or $SC$ (poly-time, polylog-space) computable multifunctions whose output is bound by a term in $L_1$ and that have access to a witness oracle for another restriction on the $Σ^{\mathsf b}_{i}$ formulas, $SIT_{i,1}$. For the $i=1$ cases, this simplifies respectively to the functions in logspace and $SC$, Steve's Class, poly-time, polylog-space. We prove independence results related to the Matiyasevich Robinson Davis Putnam Theorem (MRDP) and to whether our theories prove simultaneous nondeterministic polynomial time, sublinear space is equal to co-nondeterministic polynomial time, sublinear space.

cs.LO

MRI-GAN: A Generalized Approach to Detect DeepFakes using Perceptual Image Assessment

DeepFakes are synthetic videos generated by swapping a face of an original image with the face of somebody else. In this paper, we describe our work to develop general, deep learning-based models to classify DeepFake content. We propose a novel framework for using Generative Adversarial Network (GAN)-based models, we call MRI-GAN, that utilizes perceptual differences in images to detect synthesized videos. We test our MRI-GAN approach and a plain-frames-based model using the DeepFake Detection Challenge Dataset. Our plain frames-based-model achieves 91% test accuracy and a model which uses our MRI-GAN framework with Structural Similarity Index Measurement (SSIM) for the perceptual differences achieves 74% test accuracy. The results of MRI-GAN are preliminary and may be improved further by modifying the choice of loss function, tuning hyper-parameters, or by using a more advanced perceptual similarity metric.

cs.CV

Alternating Hierarchies for Time-Space Tradeoffs

Nepomnjascii's Theorem states that for all 0 <= ε< 1 and k > 0 the class of languages recognized in nondeterministic time n^k and space n^ε, NTISP[n^k, n^ε], is contained in the linear time hierarchy. By considering restrictions on the size of the universal quantifiers in the linear time hierarchy, this paper refines Nepomnjascii's result to give a sub- hierarchy, Eu-LinH, of the linear time hierarchy that is contained in NP and which contains NTISP[n^k, n^ε]. Hence, Eu-LinH contains NL and SC. This paper investigates basic structural properties of Eu-LinH. Then the relationships between Eu-LinH and the classes NL, SC, and NP are considered to see if they can shed light on the NL = NP or SC = NP questions. Finally, a new hierarchy, zeta -LinH, is defined to reduce the space requirements needed for the upper bound on Eu-LinH.

cs.CC

The weak pigeonhole principle for function classes in S^1_2

It is well known that S^1_2 cannot prove the injective weak pigeonhole principle for polynomial time functions unless RSA is insecure. In this note we investigate the provability of the surjective (dual) weak pigeonhole principle in S^1_2 for provably weaker function classes.

cs.LO

Quantum and Stochastic Branching Programs of Bounded Width

In this paper we show that one qubit polynomial time computations are at least as powerful as $\NC^1$ circuits. More precisely, we define syntactic models for quantum and stochastic branching programs of bounded width and prove upper and lower bounds on their power. We show any $\NC^1$ language can be accepted exactly by a width-2 quantum branching program of polynomial length, in contrast to the classical case where width 5 is necessary unless $\NC^1=\ACC$. This separates width-2 quantum programs from width-2 doubly stochastic programs as we show the latter cannot compute the middle bit of multiplication. Finally, we show that bounded-width quantum and stochastic programs can be simulated by classical programs of larger but bounded width, and thus are in $\NC^1$. The change in the revised version is the addition of the syntactic condition.

quant-ph