arXiv subjects

Da-Zheng Feng

Publications and source records attributed to Da-Zheng Feng.

5 recordsLinked to original sources

arXiv2026

Brain Learning Principles Utilizing Non-Ideal Factors in Neural Circuits

The human brain achieves its remarkable computational prowess not despite its inherent non-ideal factors noise, heterogeneity, structural irregularities, decentralized plasticity, systematic errors, and chaotic dynamics but precisely because of them. This paper systematically demonstrates that these traits, long dismissed as imperfections in classical neuroscience and eliminated in digital engineering, are evolutionary design principles that endow the brain with robustness, adaptability, and creativity.

q-bio.NC↗

arXiv2026

Learning Principles for Overcoming Non-ideal Factors in Brain

The human brain's computational prowess emerges not despite but because of its inherent "non-ideal factors"-noise, heterogeneity, structural irregularities, decentralized plasticity, systemic errors, and chaotic dynamics-challenging classical neuroscience's idealized models. These traits, long dismissed as flaws, are evolutionary adaptations that endow the brain with robustness, creativity, and adaptability. Classical frameworks falter under the brain's complexity: simulating 86 billion neurons and 100 trillion synapses is intractable, stochastic neurotransmitter release confounds signal interpretation, and the absence of global idealized models invalidates deterministic learning frameworks. Technological gaps further obscure whole-brain dynamics, revealing a disconnect between biological reality and computational abstraction.

q-bio.NC↗

arXiv2020

Fast Coherent Point Drift

Nonrigid point set registration is widely applied in the tasks of computer vision and pattern recognition. Coherent point drift (CPD) is a classical method for nonrigid point set registration. However, to solve spatial transformation functions, CPD has to compute inversion of a M*M matrix per iteration with time complexity O(M3). By introducing a simple corresponding constraint, we develop a fast implementation of CPD. The most advantage of our method is to avoid matrix-inverse operation. Before the iteration begins, our method requires to take eigenvalue decomposition of a M*M matrix once. After iteration begins, our method only needs to update a diagonal matrix with linear computational complexity, and perform matrix multiplication operation with time complexity approximately O(M2) in each iteration. Besides, our method can be further accelerated by the low-rank matrix approximation. Experimental results in 3D point cloud data show that our method can significantly reduce computation burden of the registration process, and keep comparable performance with CPD on accuracy.

cs.CV↗

arXiv2015

Human Behavior Algorithms for Highly Efficient Global Optimization

The global optimization have the very extensive applications in econometrics, science and engineering. However, the global optimization for non-convex objective functions is particularly difficult since most of the existing global optimization methods depend on the local linear search algorithms that easily traps into a local point, or the random search strategies that may frequently not produce good off-springs. According to human behavior, a one-dimensional global search method in the global optimization should adopt alternating descent and ascent (up-hill and down hill) strategies. This paper proposes the human behavior algorithms (HBA) based on alternating descent and ascent approaches along a direction or multiple different directions. Very fortunately, the proposed HBA make a global optimization method have high possibility for finding a global minimum point. Several benchmark experiments test that our HBA are highly effective for solving some benchmark optimization problems.

math.OC↗

arXiv2014

Fast Adaptive Beamforming based on kernel method under Small Sample Support

It is well-known that the high computational complexity and the insufficient samples in large-scale array signal processing restrict the real-world applications of the conventional full-dimensional adaptive beamforming (sample matrix inversion) algorithms. In this paper, we propose a computationally efficient and fast adaptive beamforming algorithm under small sample support. The proposed method is implemented by formulating the adaptive weight vector as a linear combination of training samples plus a signal steering vector, on the basis of the fact that the adaptive weight vector lies in the signal-plus-interference subspace. Consequently, by using the well-known linear kernel methods with very good small-sample performance, only a low-dimension combination vector needs to be computed instead of the high-dimension adaptive weight vector itself, which remarkably reduces the degree of freedom and the computational complexity. Experimental results validate the good performance and the computational effectiveness of the proposed methods for small samples.

cs.IT↗