arXiv subjects

Wenjun Xiao

Publications and source records attributed to Wenjun Xiao.

3 recordsLinked to original sources

arXiv2026

Reservoir Subspace Injection for Online ICA under Top-n Whitening

Reservoir expansion can improve online independent component analysis (ICA) under nonlinear mixing, yet top-$n$ whitening may discard injected features. We formalize this bottleneck as \emph{reservoir subspace injection} (RSI): injected features help only if they enter the retained eigenspace without displacing passthrough directions. RSI diagnostics (IER, SSO, $\rho_x$) identify a failure mode in our top-$n$ setting: stronger injection increases IER but crowds out passthrough energy ($\rho_x: 1.00\!\rightarrow\!0.77$), degrading SI-SDR by up to $2.2$\,dB. A guarded RSI controller preserves passthrough retention and recovers mean performance to within $0.1$\,dB of baseline $1/N$ scaling. With passthrough preserved, RE-OICA improves over vanilla online ICA by $+1.7$\,dB under nonlinear mixing and achieves positive SI-SDR$_{\mathrm{sc}}$ on the tested super-Gaussian benchmark ($+0.6$\,dB).

cs.LG↗

arXiv2026

A 1/R Law for Kurtosis Contrast in Balanced Mixtures

Kurtosis-based Independent Component Analysis (ICA) weakens in wide, balanced mixtures. We prove a sharp redundancy law: for a standardized projection with effective width $R_{\mathrm{eff}}$ (participation ratio), the population excess kurtosis obeys $|\kappa(y)|=O(\kappa_{\max}/R_{\mathrm{eff}})$, yielding the order-tight $O(c_b\kappa_{\max}/R)$ under balance (typically $c_b=O(\log R)$). As an impossibility screen, under standard finite-moment conditions for sample kurtosis estimation, surpassing the $O(1/\sqrt{T})$ estimation scale requires $R\lesssim \kappa_{\max}\sqrt{T}$. We also show that \emph{purification} -- selecting $m\!\ll\!R$ sign-consistent sources -- restores $R$-independent contrast $\Omega(1/m)$, with a simple data-driven heuristic. Synthetic experiments validate the predicted decay, the $\sqrt{T}$ crossover, and contrast recovery.

cs.LG↗

arXiv2026

A Catalog of 971 FR-I Radio Galaxies from the FIRST Survey via Hybrid Deep Learning and Ridgeline Flux Density Distribution Analysis

We present a catalog of 971 FR-I radio galaxies (FR-Is) identified from the Very Large Array Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) survey. The identifications were made using a hybrid method that combines deep learning with ridgeline flux density distribution analysis. Among these sources, 845 are new discoveries. The catalog comprises sources characterized by edge-darkened double jets, an absence of significant bent morphology, and angular sizes ranging from 23 to 159 arcseconds. Optical and/or infrared counterparts have been identified for 813 FR-Is. Among these, the host galaxies are predominantly (88.1\%) red galaxies, with the remainder (11.9\%) being blue galaxies; notably, most blue galaxies exhibit high radio power. The FR-I sample spans a radio power range of $1.20 \times 10^{21} \leq P_{\rm 1400} \leq 3.55 \times 10^{27} \, {\rm W\,Hz}^{-1}$ at 1400 MHz and reaches redshifts up to $z = 2.307$. The host galaxies have $r$-band absolute magnitudes in the range $-24 \lesssim M_r \lesssim -20$ mag. For the 512 FR-Is with estimates, the black hole masses fall within $10^7 \lesssim M_{\rm BH} \lesssim 7.94 \times 10^9 \, M_{\odot}$. Based on optical emission-line ratios and mid-infrared colors, spectroscopic classification shows that 571 hosts are low-excitation radio galaxies (LERGs) and 59 are high-excitation radio galaxies (HERGs).

astro-ph.GA↗