arXiv subjects

Sergey Salishev

Publications and source records attributed to Sergey Salishev.

6 recordsLinked to original sources

arXiv2026

Transcript-Managed Transformers: Monotone Multi-Agent Collapse and Universality with Two Pop-Enabled Transcripts

We study transcript management for fixed, finite-precision causal Transformers. A transcript is partitioned into channels of bounded blocks. Each transition consults a fixed visible suffix and may append one block, leaving the model, weights, and token protocol unchanged. The operation $P_c:=\PopContext(c)$ deletes the newest block on channel $c$ and exposes its predecessor. We model the layer by the Transcript-Managed Transducer $\TMTn{k}$: one finite controller, $k$ channels, and per-round actions from stay, push, and pop under a caller-driven status map. Fixed visible windows encode as finite symbols. The pop-free Restricted Transcript-Managed Transducer $\RTMTn{k}$ is the standard append-only layer and, for every fixed $k$, realizes exactly the deterministic finite-state transductions. The same holds for every fixed finite agent population under a monotone protocol that appends, routes, and copies visible blocks. Admitting $\{P_c\}_{c=1}^k$ restores pop. Newest-first, a pop-enabled channel is a stack; compiling to the Hopcroft--Ullman presentation transfers the classical hierarchy: $\DCFL$ for $k=1$ and $\RE$ for every $k\ge2$. Orchestrated one-channel agents match one controller with $k$ channels, so two pop-enabled transcripts---in one agent or two---suffice for universality. Simulation costs and invariance to fixed block size and visible radius are stated. The bounds fix precision, alphabets, blocks, visibility, controller state, and population; growing exact context, hidden-block access, writable stores, and unbounded \textbf{Spawn} add further state.

cs.LG↗

arXiv2026

Local Stability and Gaussian Smoothing of Quantized Neural Networks

We study Gaussian averaging as a smooth surrogate for quantized neural models. Under bounded local oscillation, we derive a local dimension-dependent bound on |f-g|, linking Gaussian smoothing to the stability analysis of discontinuous networks. We compute closed-form Gaussian averages of the rectified linear unit (ReLU) and sign activation functions, and illustrate the mechanism on a high-dimensional binary perceptron, where layer-preactivation aggregation under an explicit quantization-noise surrogate yields the Gaussian envelope used in inference-side smoothing and training-side smooth surrogate gradients.

cs.LG↗

arXiv2025

Synthesis of signal processing algorithms with constraints on minimal parallelism and memory space

This thesis develops signal-processing algorithms and implementation schemes under constraints of minimal parallelism and memory space, with the goal of improving energy efficiency of low-power computing hardware. We propose (i) a power/energy consumption model for clocked CMOS logic that supports selecting optimal parallelism, (ii) integer-friendly approximation methods for elementary functions that reduce lookup-table size via constrained piecewise-polynomial (quasi-spline) constructions with accuracy guarantees, (iii) provably conflict-free data placement and execution order for mixed-radix streaming FFT on multi-bank and single-port memories, including a self-sorting FFT variant, and (iv) a parallelism/memory analysis of the fast Schur algorithm for superfast Toeplitz system solving, motivated by echo-cancellation workloads. The results provide constructive theorems, schedules, and design trade-offs enabling efficient specialized accelerators.

eess.SP↗

arXiv2025

The Narrow Corridor of Stable Solutions in an Extended Osipov--Lanchester Model with Constant Total Population

This paper considers a modification of the classical Osipov--Lanchester model in which the total population of the two forces $N=R+B$ is preserved over time. It is shown that the dynamics of the ratio $y=R/B$ reduce to the Riccati equation $\dot y=αy^2-β$, which admits a complete analytical study. The main result is that asymptotically stable invariant sets in the positive quadrant $R,B\ge 0$ exist exactly in three sign cases of $(α,β)$: (i) $α<0,β<0$ (stable interior equilibrium), (ii) $α=0,β<0$ (the face $B=0$ is stable), (iii) $α<0,β=0$ (the face $R=0$ is stable). For $α>0$ or $β>0$ the solutions reach the boundaries of applicability of the model in finite time. Moreover, $α<0,β<0$ corresponds to exponential growth of solutions in the original system. Passing to a model perturbed in $α(t),β(t)$ requires buffer dynamics repelling from the axes to preserve stability of the solution.

math.DS↗

arXiv2025

GDNSQ: Gradual Differentiable Noise Scale Quantization for Low-bit Neural Networks

Quantized neural networks can be viewed as a chain of noisy channels, where rounding in each layer reduces capacity as bit-width shrinks; the floating-point (FP) checkpoint sets the maximum input rate. We track capacity dynamics as the average bit-width decreases and identify resulting quantization bottlenecks by casting fine-tuning as a smooth, constrained optimization problem. Our approach employs a fully differentiable Straight-Through Estimator (STE) with learnable bit-width, noise scale and clamp bounds, and enforces a target bit-width via an exterior-point penalty; mild metric smoothing (via distillation) stabilizes training. Despite its simplicity, the method attains competitive accuracy down to the extreme W1A1 setting while retaining the efficiency of STE.

cs.LG↗

arXiv2025

Regular mixed-radix DFT matrix factorization for in-place FFT accelerators

The generic vector memory based accelerator is considered which supports DIT and DIF FFT with fixed datapath. The regular mixed-radix factorization of the DFT matrix coherent with the accelerator architecture is proposed and the correction proof is presented. It allows better understanding of architecture requirements and simplifies the developing and proving correctness of more complicated algorithms and conflict-free addressing schemes.

cs.AR↗