vishalv.com

Research notes on The past, present and future of de novo protein design.

We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early ’90s, where a slow neural net learns by gradient descent to program the fast weights of another net through sequences of elementary programmi…

"This book covers modern deep learning and tackles supervised learning, model architecture, unsupervised learning, and deep reinforcement learning"--

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window.

Psychostimulants act on monoamine systems including the dopamine transporter (DAT), a key regulator of dopamine reuptake following synaptic release.

This paper proposes that a fundamental organizational trend in biological, cognitive, cultural, and technological evolution is the emergence of progressively more capable representation–interpretation systems.

Modern AI systems are increasingly being evaluated for their ability to reason, code, prove theorems, use tools, and long-horizon research tasks.

What does it mean to understand the world?

A growing body of work points to the great promise of AI systems that can continually expand their capabilities as they operate in an open-ended environment.

Research notes on Elements of information theory.

Although most theories posit that natural behavior can be explained as maximizing some form of extrinsic reward, often called utility, some behaviors appear to be reward independent.

In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internet-scale data.

Deep networks learn internal representations whose geometry—how features bend, rotate, and evolve—affects both generalization and robustness.

The Platonic Representation Hypothesis suggests that representations from neural networks are converging to a common statistical model of reality.

Vision-Language-Action (VLA) models have achieved strong semantic generalization for embodied policy learning, yet they learn reactive observation-to-action mappings without explicitly modeling how the physical world evolves under intervention.

Bayesian inference typically focuses upon two issues.

A long-standing goal in robot learning is to develop methods for robots to acquire new skills autonomously.

Research notes on Algebraic topology.

Progress in embodied intelligence increasingly depends on scalable data infrastructure.

Research notes on CS168: The Modern Algorithmic Toolbox Lectures #11: Spectral Graph Theory, I.

Probabilistic graphical modeling is a branch of machine learning that uses probability distributions to describe the world, make predictions, and support decision-making under uncertainty.

Research notes on DISCRETE DIFFERENTIAL GEOMETRY: AN APPLIED INTRODUCTION.

Drawing on the correspondence between the graph Laplacian, the Laplace-Beltrami operator on a manifold , and the connections to the heat equation , we propose a geometrically motivated algorithm for constructing a representation for data sampled from a low

This tutorial serves as a comprehensive guide for understanding graph databases, focusing on the fundamentals of graph theory while showcasing practical applications across various fields.

The Hopf fibration is an important object in mathematics and physics.

Post-GPU compute is a bet that new hardware will make different primitives cheap.

Connecting probabilistic graphical models to energy-based formulations, and showing how sampling replaces optimization as the core mechanism for inference and computation.

System design evaluation of the Zotero iOS client based on long-term usage as a knowledge power user tool.

I finally deleted my WordPress setup and replaced it with a system I actually control.