Yannis Psaromiligkos Paper Accepted in IEEE Wireless Communications

Massive upgrade to my cosmological framework (Specular Bit Architecture) to align it with strict peer-review standards. If you’ve read the previous versions, here are the absolute core upgrades in this latest release: 1️⃣ CPT Symmetry Formally Integrated: The new manuscript explicitly proves how Charge, Parity, and Time reversal symmetries are mathematically covariant within the discrete digital logic of the SBA information action (Theorem 7.2). 2️⃣ Strict Epistemological Separation: I’ve implemented a "Layer-Tagged Architecture" to enforce a hard boundary between formal/mathematical proofs (theorems, exact local rewrites) and hypothesis/modeling layers (like DNA mapping and deep-space metrics). 3️⃣ Reproducible Python Proofbench: The mathematical claims are now backed by a rigorous Python validation engine. Anyone can run the script to exhaustively generate the 3,125 local neighborhood states and extract the bounded-propagation constants automatically. 4️⃣ Cryptographic Data Integrity: To ensure 100% transparent Open Science, the Proofbench now generates SHA-256 hashes for all computational certificates and artifacts. The framework is now mathematically armored and ready for journal submission. https://lnkd.in/d9DkDa_G 🌌🧬💻 #SBA #TheoreticalPhysics #CPTSymmetry #Cosmology #OpenScience #PeerReview #Python

I am pleased to announce the publication of my first arXiv preprint: "Accelerating Quantum State Encoding with SIMD: Design, Implementation, and Benchmarking." Efficient data encoding has long been a bottleneck in hybrid quantum-classical algorithms, with traditional simulators spending significant time converting classical features into quantum rotations. To address this, our team developed the Hybriqu Encoder, a Rust-based, SIMD-aware kernel that focuses exclusively on angle encoding and integrates seamlessly with Python via CFFI. By processing multiple double-precision rotations simultaneously using AVX-class vector lanes, utilizing pre-calculated trigonometric factors, and optimizing for cache performance, we achieved a 5.4% speedup at 64 qubits on Apple Silicon. This performance scales effectively as data exceeds the L1 cache size, demonstrating that vectorization provides consistent improvements for compute-bound encoding tasks. Our benchmarking also identified that kernels applying rotations to the entire state vector remain heavily limited by memory bandwidth rather than compute power. This highlights a clear path for future optimization regarding state updates and processing methods in larger hybrid quantum-classical systems. A huge thank you to my co-authors Irwan Alnarus Kautsar, Gunawan Witjaksono, and Haza Nuzly Bin Abdull Hamed for their collaboration and support on this work. You can read the full preprint here: https://lnkd.in/gKMNt57B #QuantumComputing #MachineLearning #RustLang #HighPerformanceComputing #SIMD #QuantumSimulation #ArXiv

🔔 𝐄𝐱𝐜𝐢𝐭𝐢𝐧𝐠 𝐔𝐩𝐝𝐚𝐭𝐞𝐬 𝐨𝐧 “𝐊𝐊𝐓-𝐇𝐚𝐫𝐝𝐍𝐞𝐭” 🔔 𝐊𝐊𝐓-𝐇𝐚𝐫𝐝𝐍𝐞𝐭 —a general physics-constrained ML tool that combines data and domain knowledge for scientific machine learning (SciML)—is now officially available as a Python package. We’ve significantly upgraded the framework with CUDA support and a more modular problem construction pipeline, making it faster and easier to use than before. 𝑲𝒆𝒚 𝒊𝒎𝒑𝒓𝒐𝒗𝒆𝒎𝒆𝒏𝒕𝒔: - 📈 𝘐𝘮𝘱𝘳𝘰𝘷𝘦𝘥 𝘱𝘳𝘦𝘥𝘪𝘤𝘵𝘪𝘰𝘯 𝘢𝘤𝘤𝘶𝘳𝘢𝘤𝘺 𝘢𝘯𝘥 𝘪𝘯𝘧𝘦𝘳𝘦𝘯𝘤𝘦 𝘵𝘪𝘮𝘦 - ⏱️ 𝘍𝘢𝘴𝘵𝘦𝘳 𝘢𝘯𝘥 𝘮𝘰𝘳𝘦 𝘦𝘧𝘧𝘪𝘤𝘪𝘦𝘯𝘵 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨 𝘰𝘯 𝘣𝘰𝘵𝘩 𝘊𝘗𝘜 𝘢𝘯𝘥 𝘎𝘗𝘜 - 🧩 𝘔𝘰𝘥𝘶𝘭𝘢𝘳 𝘥𝘦𝘴𝘪𝘨𝘯 𝘧𝘰𝘳 𝘧𝘭𝘦𝘹𝘪𝘣𝘭𝘦 𝘱𝘳𝘰𝘣𝘭𝘦𝘮 𝘴𝘦𝘵𝘶𝘱 𝑰𝒇 𝒚𝒐𝒖’𝒗𝒆 𝒃𝒆𝒆𝒏 𝒖𝒔𝒊𝒏𝒈 𝒆𝒂𝒓𝒍𝒊𝒆𝒓 𝒗𝒆𝒓𝒔𝒊𝒐𝒏𝒔, 𝒘𝒆 𝒔𝒕𝒓𝒐𝒏𝒈𝒍𝒚 𝒓𝒆𝒄𝒐𝒎𝒎𝒆𝒏𝒅 𝒔𝒘𝒊𝒕𝒄𝒉𝒊𝒏𝒈 𝒕𝒐 𝒕𝒉𝒊𝒔 𝒐𝒑𝒕𝒊𝒎𝒊𝒛𝒆𝒅 𝒊𝒎𝒑𝒍𝒆𝒎𝒆𝒏𝒕𝒂𝒕𝒊𝒐𝒏. 📄 Paper: https://lnkd.in/gD7p7G6Z 💻 Code: https://lnkd.in/gzqrEVgf 📦 Package: https://lnkd.in/gNrDxF3t ⚙️ Install via pip: CPU: 𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘬𝘬𝘵-𝘩𝘢𝘳𝘥𝘯𝘦𝘵 GPU (CUDA 12): 𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 "𝘬𝘬𝘵-𝘩𝘢𝘳𝘥𝘯𝘦𝘵[𝘤𝘶𝘥𝘢12]" We’d love to hear your feedback. Feel free to reach out with questions or thoughts on the documentation and examples. Bimol Nath Roy Rahul Golder Ashfaq Iftakher #PhysicsConstrainedMachineLearning #PCML #ConstrainedLearning #MachineLearning #Optimization #DeepLearning #JAX #CUDA #Research #AI #Engineering #PSE

⚛️ From Quantum Mechanics Foundations to Chip-Scale Maser Understanding quantum behavior in real devices is complex. Whether it's a two-level system (|g⟩ & |e⟩) in a diamond NV-center or population inversion in a maser — the physics starts from the same place: Quantum mechanics. But visualizing this in real-time? Almost impossible with static diagrams. 💡 Our Solution: My collaborator Shahariar Ryehan and I built a Python-powered animated simulation that walks through every key concept — from wavefunction to working maser. 🔑 Key Insights: → Two-Level System: Energy states |g⟩ (ground) in BLUE and |e⟩ (excited) in RED — the heart of every maser → Wavefunction Probability: dP = |ψ(r,t)|² dV — where the quantum particle is likely to be found → Density Matrix: ρ̂ = [[ρₑₑ, ρₑ₉], [ρ₉ₑ, ρ₉₉]] — handles mixed states in real experiments → Boltzmann Distribution: At room temperature (300K), Nₑ/N₉ ≈ 0.9995 — NO maser! → Population Inversion: Pump laser makes Nₑ > N₉ — maser condition achieved! → Chip Challenge: Surface defects cause T₁ relaxation — energy leaks out → The Solution: Fast optical pumping (532nm green laser) with P_pump > 1/T₁ → Final Result: When condition met → ρₑₑ > ρ₉₉ → MASER OSCILLATES! 🛠️ Built with: Python, Matplotlib, NumPy, FFMpeg 🔗 Run the code & watch the animation: https://lnkd.in/gpm6wtgF https://lnkd.in/g7yUc_EH Huge thanks to my collaborator Shahariar Ryehan for numerical implementation & animation design. #QuantumMechanics #Maser #DiamondNV #PopulationInversion #Python #Simulation #TwoLevelSystem #DensityMatrix #BoltzmannDistribution #ChipScaleMaser

🔥🧬 𝐒𝐭𝐞𝐫𝐞𝐨𝐌𝐨𝐥𝐆𝐫𝐚𝐩𝐡: 𝐒𝐭𝐞𝐫𝐞𝐨𝐜𝐡𝐞𝐦𝐢𝐬𝐭𝐫𝐲-𝐀𝐰𝐚𝐫𝐞 𝐌𝐨𝐥𝐞𝐜𝐮𝐥𝐚𝐫 & 𝐑𝐞𝐚𝐜𝐭𝐢𝐨𝐧 𝐆𝐫𝐚𝐩𝐡𝐬 ⚡🔗 ✅ StereoMolGraph is an open-source Python library for stereochemistry-aware molecular and reaction graph representations. ✅𝘊𝘰𝘯𝘷𝘦𝘯𝘵𝘪𝘰𝘯𝘢𝘭 𝘨𝘳𝘢𝘱𝘩𝘴 𝘰𝘧𝘵𝘦𝘯 𝘧𝘢𝘪𝘭 𝘵𝘰 𝘦𝘯𝘤𝘰𝘥𝘦 𝘴𝘵𝘦𝘳𝘦𝘰𝘤𝘩𝘦𝘮𝘪𝘴𝘵𝘳𝘺, particularly in symmetric molecules, non-tetrahedral centers, and transition states. ✅The library employs permutation-invariant local stereodescriptors, grounded in group theory, to robustly capture chirality. ✅This enables reliable comparison of stereoisomers, identification of enantiomerism and diastereomerism, and support for fleeting stereochemistry in transition states. ✅With RDKit interoperability and visualization, StereoMolGraph provides a transparent, extensible tool for analyzing complex organic molecules, metal complexes, and distinct chiral reaction pathways in advanced chemoinformatics workflows. Source: Maxim PapushaKai Leonhard. https://lnkd.in/gX_SBYcD

Day 25/30; Transformer Cross-Attention bridge; Problem : PGN-to-FEN Translation. Mapping a 1D sequence of events (chess move history) into a static 2D spatial snapshot (board layout). Architecture Implemented : 1. Zero Recurrence : Replaced the O(n) sequential bottleneck with O(1) parallel matrix multiplication using Scaled Dot-Product Attention. 2. Positional Encoding : Injected sine/cosine frequencies directly into the token embeddings to mathematically plot time geographically. 3. Multi-Head Bridge : The Decoder (Space) fires a Query (Q) at the Encoder's historical Keys (K), computing a dot-product alignment score to extract the exact Value (V) payload of the move. 4. VRAM Optimization : Aggressively cropped the context window and implemented bitwise padding + causal masking to prevent future-state leakage during autoregressive training. Genuinely one of the most interesting problems I have solved; I thank god for letting me have the ability to experience this beauty of a model and learn it. I'm gonna call this the QKV Gambit 😆 . The math is just beautiful. The model doesn’t "memorize" boards; it physically learns how to simulate piece trajectories across an attention matrix. Explanation, Math, Pytorch Implementation in the repository. Repo : https://lnkd.in/g8swe4yC #MachineLearning #DeepLearning #Transformers #AI #Python #DataScience #SoftwareEngineering #30DayChallenge

What is the first paper you should read about stochastic optimization? Some might suggest one about training neural networks, but I disagree. Stochastic optimization is much more than empirical risk minimization. It addresses the deeper question: how to make a good decision now while thinking about possible recourse decisions later when we have more information. From this perspective, an excellent starting point is Higle & Wallace, "Sensitivity analysis and uncertainty in linear programming," Interfaces, 2003 https://lnkd.in/g-Fypw5F

Vision Transformers using Pytorch trained a small Vision Transformer architecture from scratch using PyTorch to better understand how they work. The 28 x 28 input image is converted into a 4 x 4 grid of 7 x 7 patches. This is done using a convolutional operation, resulting in a sequence shape of (16, 500) (16 patches with an embedding dimension of 500). Next, we apply positional embeddings via element-wise addition, so the shape remains the same. We then calculate the Queries (Q), Keys (K), and Values (V) by matrix-multiplying the (16, 500) sequence with the learnable 500 x 500 weight matrices Wq , WK, and Wv. From here, we calculate the attention scores, apply a Softmax, and multiply by V to get our (16, 500) tensor. Finally, we apply layer normalization, pass the features through a feed-forward network to expand and contract the dimensions, and route the output to a linear classifier. To test the pipeline, I trained it on the MNIST dataset and achieved ~94% accuracy in 5 epochs #ComputerVision #DeepLearning #VisionTransformers

🧠 Built a Vanilla GAN from scratch using PyTorch — and it was one of the most rewarding deep learning projects I've done! Generative Adversarial Networks always felt like "black magic" to me — until I actually built one. So I rolled up my sleeves and implemented everything from the ground up, no shortcuts. 🔧 Here's what's under the hood: • Dataset: CelebA (celebrity faces), preprocessed with center crop → resize to 64×64 → normalized to [-1, 1] • Generator: 4-layer MLP (noise z=100 → 256 → 512 → 1024 → image), Tanh output • Discriminator: 4-layer MLP with LeakyReLU activations + Sigmoid output • Loss: Binary Cross-Entropy (BCELoss) • Optimizer: Adam (lr=0.0002, β₁=0.5) for both networks • Hardware: Auto-selects CUDA / Apple MPS / CPU The core idea is elegant — two networks locked in a constant game. The Generator learns to create realistic faces from pure noise, while the Discriminator learns to tell real from fake. Over 5 epochs, you can watch the generated faces slowly come to life. 🎨 If you're learning deep learning, building a GAN from scratch will teach you more about backprop and optimization than almost anything else. 🔗 GitHub: https://lnkd.in/gVV4FXz5 #DeepLearning #GAN #PyTorch #GenerativeAI #MachineLearning #ComputerVision #Python

Day 24/30 ML Challenge; Bahdanau Attention; Standard Encoder-Decoder architectures suffer from catastrophic amnesia because they force the Encoder to compress entire sequences into a single fixed-size vector. To solve this, I engineered an Attention bridge that dynamically calculates alignment scores, allowing the Decoder to "look back" at specific Encoder hidden states at every single generation step. Core Mechanics; 1. Architecture : A Bidirectional GRU Encoder paired with an Attention-driven Unidirectional GRU Decoder. 2. Tokenization : Strict character-level mapping to prevent the infinite vocabulary explosion inherent to mathematical domains. 3. Evaluation : Exact Match Accuracy (EMA). Character Error Rate is useless in calculus; a single hallucinated token invalidates the entire equation. 4. Data Pipeline : Engineered a deterministic synthetic generator using SymPy to build abstract syntax trees and exact ground-truth targets. The architecture works, but the mathematical engine is too slow to scale. Full Explanation, Math and Python Code in Repository. Repo : https://lnkd.in/gj-pd8dg #MachineLearning #PyTorch #DeepLearning #ArtificialIntelligence #SequenceModeling #Engineering #DataScience #AI