I’ve been exploring the “Inference Paradox”—the performance gap between transcendental-heavy activations (Swish/GELU) and hardware-efficient but jagged approximations (HardSwish). I am sharing SATIN-U (Smoothstep-Activated Trainable Inference Network), which utilizes a cubic polynomial bridge to achieve Swish-like fidelity without the exponential math tax. The Implementation Logic: The goal was to maintain a differentiable path while ensuring an absolute zero floor for hardware-level sparsity (clock gating). The Math: u = clamp(0.5 + 0.5 * (x / b), 0, 1) gate = […]

Ver mais

Traditional mean–variance portfolio optimization proves inadequate for cryptocurrency markets, where extreme volatility, fat-tailed return distributions, and unstable correlation structures undermine the validity of variance as a comprehensive risk measure. To address these limitations, this paper proposes a unified entropy-based portfolio optimization framework grounded in the Maximum Entropy Principle (MaxEnt). Within this setting, Shannon entropy, Tsallis entropy, and Weighted Shannon Entropy (WSE) are formally derived as particular specifications of a common constrained optimization problem solved via the method of […]

Ver mais

In the framework of (fuzzy) Multi-Criteria Decision-Making, we propose a method that1 allows the decision maker to subjectively approach the problem by suitably modifying the decision matrix. We consider a decision problem related to a random quantity X with set of values {x1, x2, . . . , xn}, and a set of properties {C1, C2, . . . , Cm}of X. In this setting, the properties Cj are the criteria of the decision problem, the alternatives represent […]

Ver mais

Conventional semi-supervised contrastive learning methods assign pseudo-labels only to samples whose highest predicted class probability exceeds a predefined threshold, and then perform supervised contrastive learning using those selected samples. In this study, we propose a novel loss function that estimates the confidence of each sample based on the entropy of its predicted probability distribution and applies confidence-based adaptive weighting. This approach enables pseudo-label assignment even to samples that were previously excluded from training and facilitates contrastive learning that […]

Ver mais

Goldbach’s conjecture, one of the oldest and most resilient problems in number theory, has traditionally been approached through additive and combinatorial methods. Despite extensive numerical verification and partial results, a structural explanation for its apparent universality remains elusive. In this work, we propose a reinterpretation of Goldbach’s conjecture within the framework of Viscous Time Theory (VTT), introducing an informational–geometric perspective in which prime numbers are treated as stable coherence attractors in an informational field. Within this framework, the […]

Ver mais

The constrained combinatorial multi-armed bandit model has been widely employed to solve problems in wireless networking and related areas, including the problem of wireless scheduling for throughput optimization under unknown channel conditions. Most work in this area uses an algorithm design strategy that combines a bandit learning algorithm with the virtual queue technique to track the throughput constraint violation. These algorithms seek to minimize the virtual queue length in their algorithm design. However, in networks where channel conditions […]

Ver mais

I’ve been learning to code my whole life. I’ve mastered several programming languages, including C++, Delphi, C#, JavaScript, Java, and Go. I’ve created numerous software projects in different domains. Learning is always part of any programming language, platform, or framework. The learning process has changed over the past 20 years. I used to spend a lot of time searching for good books on C++ or Delphi in a bookstore. Then I’d googled problems that weren’t covered in the […]

Ver mais

arXiv:2601.02451v1 Announce Type: new Abstract: Graph Neural Networks (GNNs) suffer from over-smoothing in deep architectures and expressiveness bounded by the 1-Weisfeiler-Leman (1-WL) test. We adapt Manifold-Constrained Hyper-Connections (mhc)~citep{xie2025mhc}, recently proposed for Transformers, to graph neural networks. Our method, mHC-GNN, expands node representations across $n$ parallel streams and constrains stream-mixing matrices to the Birkhoff polytope via Sinkhorn-Knopp normalization. We prove that mHC-GNN exhibits exponentially slower over-smoothing (rate $(1-gamma)^{L/n}$ vs. $(1-gamma)^L$) and can distinguish graphs beyond 1-WL. Experiments on 10 […]

Ver mais