DEV Community

Umesh Tharuka Malaviarachchi profile picture

Umesh Tharuka Malaviarachchi

Founder & CEO at Histic | Social Media Manager ♾️ | Microsoft Certified Advertising Professional | Meta Certified Digital Marketing Associate | Srilanka's 1st LinkedIn Certified Marketing Insider

Location Colombo, Sri Lanka Joined Joined on  Personal website https://umeshtharukaofficial.hashnode.dev/ github website

Education

Cisco Certified

Model Soups: Why Averaging Fine-Tuned Weights Beats Picking the Best One

Model Soups: Why Averaging Fine-Tuned Weights Beats Picking the Best One

Comments
7 min read
Amortised Inference: Why Training a Network to Infer Beats Running Optimisation Every Time

Amortised Inference: Why Training a Network to Infer Beats Running Optimisation Every Time

Comments
7 min read
Proximal Methods: Why Splitting a Nonsmooth Problem Beats Solving It Directly

Proximal Methods: Why Splitting a Nonsmooth Problem Beats Solving It Directly

Comments
7 min read
Distributional Reinforcement Learning: Why Predicting a Distribution Beats Predicting a Mean

Distributional Reinforcement Learning: Why Predicting a Distribution Beats Predicting a Mean

Comments
10 min read
Causal Representation Learning: Why Observational Data Alone Cannot Identify Latent Causes

Causal Representation Learning: Why Observational Data Alone Cannot Identify Latent Causes

Comments
8 min read
Neural Ordinary Differential Equations: Why Depth Can Be a Continuous Variable

Neural Ordinary Differential Equations: Why Depth Can Be a Continuous Variable

Comments
8 min read
Sparse Autoencoders: Why Forcing Sparsity Makes Neural Features Interpretable

Sparse Autoencoders: Why Forcing Sparsity Makes Neural Features Interpretable

Comments
8 min read
Contrastive Learning: Why Pulling Similar Points Together Learns Representations Without Labels

Contrastive Learning: Why Pulling Similar Points Together Learns Representations Without Labels

Comments
8 min read
Kernel Mean Embeddings: Turning Distributions into Points You Can Compute With

Kernel Mean Embeddings: Turning Distributions into Points You Can Compute With

Comments
7 min read
Bilevel Optimisation: Why Meta-Learning Is an Optimisation Problem Inside an Optimisation Problem

Bilevel Optimisation: Why Meta-Learning Is an Optimisation Problem Inside an Optimisation Problem

Comments
9 min read
Bayesian Optimization: Why a Surrogate Model Plus an Acquisition Function Beats Grid Search

Bayesian Optimization: Why a Surrogate Model Plus an Acquisition Function Beats Grid Search

Comments
10 min read
Diffusion Models: How Reversing Noise Learns a Data Distribution

Diffusion Models: How Reversing Noise Learns a Data Distribution

Comments
8 min read
Thompson Sampling: Why Randomised Exploration Beats Greedy Optimisation

Thompson Sampling: Why Randomised Exploration Beats Greedy Optimisation

Comments
7 min read
Causal Discovery: Why Conditional Independence Can Recover a Graph - and When It Cannot

Causal Discovery: Why Conditional Independence Can Recover a Graph - and When It Cannot

Comments
7 min read
Low-Rank Adaptation: Why Freezing Most Weights Still Fine-Tunes a Large Model

Low-Rank Adaptation: Why Freezing Most Weights Still Fine-Tunes a Large Model

Comments
8 min read
State Space Models: Why Linear Recurrences Rival Attention for Long Sequences

State Space Models: Why Linear Recurrences Rival Attention for Long Sequences

Comments
7 min read
Evaluating Retrieval-Augmented Generation: Why End-to-End Accuracy Hides the Real Failures

Evaluating Retrieval-Augmented Generation: Why End-to-End Accuracy Hides the Real Failures

Comments
9 min read
Neural Collapse: Why the Last Layer Converges to a Simple Geometric Structure

Neural Collapse: Why the Last Layer Converges to a Simple Geometric Structure

Comments
8 min read
Rotary Position Embeddings: How Rotating Queries and Keys Encodes Relative Position

Rotary Position Embeddings: How Rotating Queries and Keys Encodes Relative Position

Comments
7 min read
In-Context Learning: How a Forward Pass Learns Without Updating a Single Weight

In-Context Learning: How a Forward Pass Learns Without Updating a Single Weight

Comments
10 min read
Normalizing Flows: Why Invertible Transformations Make Exact Likelihoods Possible

Normalizing Flows: Why Invertible Transformations Make Exact Likelihoods Possible

Comments
7 min read
The Lottery Ticket Hypothesis: Why Sparse Subnetworks Train as Well as Dense Ones

The Lottery Ticket Hypothesis: Why Sparse Subnetworks Train as Well as Dense Ones

Comments
7 min read
Energy-Based Models: Why an Unnormalised Density Can Still Define a Distribution

Energy-Based Models: Why an Unnormalised Density Can Still Define a Distribution

Comments
10 min read
Variational Inference: Why the ELBO Trades Accuracy for Tractability

Variational Inference: Why the ELBO Trades Accuracy for Tractability

Comments
7 min read
Gaussian Processes: Why Kernels Are Priors Over Functions, Not Just Similarity Scores

Gaussian Processes: Why Kernels Are Priors Over Functions, Not Just Similarity Scores

Comments
8 min read
Direct Preference Optimization: Why You Can Align a Model Without a Reward Model

Direct Preference Optimization: Why You Can Align a Model Without a Reward Model

Comments
8 min read
Mixture of Experts: Why Sparse Routing Scales Models Without Scaling Compute

Mixture of Experts: Why Sparse Routing Scales Models Without Scaling Compute

Comments
7 min read
Double Descent: Why Bigger Models Can Generalise Better After Getting Worse

Double Descent: Why Bigger Models Can Generalise Better After Getting Worse

Comments
7 min read
Sharpness-Aware Minimization: Why Flat Minima Generalise Better Than Sharp Ones

Sharpness-Aware Minimization: Why Flat Minima Generalise Better Than Sharp Ones

Comments
8 min read
Grokking: Why Neural Networks Suddenly Generalise Long After Memorising

Grokking: Why Neural Networks Suddenly Generalise Long After Memorising

Comments
9 min read
Federated Learning: Why Averaging Model Updates Works - and When It Fails

Federated Learning: Why Averaging Model Updates Works - and When It Fails

Comments
8 min read
Graph Neural Networks: Why Message Passing Works - and Where It Breaks

Graph Neural Networks: Why Message Passing Works - and Where It Breaks

Comments
9 min read
Hamiltonian Monte Carlo: How Momentum Turns Random Walks into Efficient Sampling

Hamiltonian Monte Carlo: How Momentum Turns Random Walks into Efficient Sampling

1
Comments
8 min read
Score Matching: Why Learning the Gradient of a Log-Density Beats Learning the Density Itself

Score Matching: Why Learning the Gradient of a Log-Density Beats Learning the Density Itself

Comments
8 min read
Stein Variational Gradient Descent: Particles That Repel Their Way to the Posterior

Stein Variational Gradient Descent: Particles That Repel Their Way to the Posterior

Comments
9 min read
Optimal Transport: Why Wasserstein Distance Sees Geometry That KL Divergence Misses

Optimal Transport: Why Wasserstein Distance Sees Geometry That KL Divergence Misses

Comments
9 min read
The Information Bottleneck: Why the Best Representations Forget Most of What They See

The Information Bottleneck: Why the Best Representations Forget Most of What They See

Comments
8 min read
Conformal Prediction: Valid Uncertainty Without Distributional Assumptions

Conformal Prediction: Valid Uncertainty Without Distributional Assumptions

Comments
9 min read
Spectral Bias: Why Neural Networks Learn Low Frequencies First

Spectral Bias: Why Neural Networks Learn Low Frequencies First

Comments
7 min read
Natural Gradients: Why Steepest Descent Depends on the Geometry You Choose

Natural Gradients: Why Steepest Descent Depends on the Geometry You Choose

Comments
8 min read
Doubly Robust Estimation: When Two Imperfect Models Still Get Causality Right

Doubly Robust Estimation: When Two Imperfect Models Still Get Causality Right

Comments
8 min read
Why a simple for loop (brute-force k-NN) becomes computationally impossible for modern AI applications

Why a simple for loop (brute-force k-NN) becomes computationally impossible for modern AI applications

Comments
4 min read
Google Genie 3 AI: The Future of Game Development Has Arrived!

Google Genie 3 AI: The Future of Game Development Has Arrived!

Comments
3 min read
Dive Deeper with AI Mode: Circle to Search Now Gives You a Gamer’s Edge

Dive Deeper with AI Mode: Circle to Search Now Gives You a Gamer’s Edge

Comments
3 min read
How Do Diffusion Models Work? An Introduction to Generative Image Processing

How Do Diffusion Models Work? An Introduction to Generative Image Processing

Comments 1
2 min read
Windsurf Introduces a Free‑to‑Use SWE‑1 Coding Model Family

Windsurf Introduces a Free‑to‑Use SWE‑1 Coding Model Family

Comments
3 min read
Google Cloud’s Next-Gen AI and CJIS 6.0: A New Era for Public Safety

Google Cloud’s Next-Gen AI and CJIS 6.0: A New Era for Public Safety

Comments
2 min read
Google Cloud's Open Lakehouse: Powering the Future of AI with Open Data and Unprecedented Performance

Google Cloud's Open Lakehouse: Powering the Future of AI with Open Data and Unprecedented Performance

1
Comments
5 min read
Deep Dive into Google Agentspace's Core Architecture: Orchestrating the Enterprise AI Revolution

Deep Dive into Google Agentspace's Core Architecture: Orchestrating the Enterprise AI Revolution

13
Comments
5 min read
Unlocking Deeper Security: TLS Inspection Now Live in Microsoft Entra Internet Access!

Unlocking Deeper Security: TLS Inspection Now Live in Microsoft Entra Internet Access!

Comments
6 min read
The AI Architect's Toolkit: Navigating the Core Paradigms of Supervised vs. Unsupervised Learning

The AI Architect's Toolkit: Navigating the Core Paradigms of Supervised vs. Unsupervised Learning

Comments
9 min read
Revolutionize Your Workflow: Boosting SQL Productivity with Copilot in SSMS 21

Revolutionize Your Workflow: Boosting SQL Productivity with Copilot in SSMS 21

Comments
4 min read
Empowering Agentic AI: The Symbiotic Power of Microsoft Fabric and Azure AI Foundry

Empowering Agentic AI: The Symbiotic Power of Microsoft Fabric and Azure AI Foundry

3
Comments 1
7 min read
Welcome to Firebase Studio: Your Unified Command Center for App Development

Welcome to Firebase Studio: Your Unified Command Center for App Development

7
Comments
6 min read
Gemini Just Leveled Up: Personalized Responses Powered by Your Google Activity

Gemini Just Leveled Up: Personalized Responses Powered by Your Google Activity

2
Comments
3 min read
DeepSeek Under Siege: Cyberattack and the Paradox of Explosive AI Growth

DeepSeek Under Siege: Cyberattack and the Paradox of Explosive AI Growth

Comments
5 min read
Google's Willow Chip: Beyond the Hype, a Quantum Leap in Hardware is Here

Google's Willow Chip: Beyond the Hype, a Quantum Leap in Hardware is Here

Comments
3 min read
The Irrational Consumer: How Behavioral Science is Revolutionizing Brand Strategy

The Irrational Consumer: How Behavioral Science is Revolutionizing Brand Strategy

2
Comments 1
5 min read
Automating Business Intelligence Dashboards Using Gemini AI: A Deep Dive

Automating Business Intelligence Dashboards Using Gemini AI: A Deep Dive

2
Comments 1
7 min read
Exploring Azure AI Agent Service: A Leap in Conversational AI

Exploring Azure AI Agent Service: A Leap in Conversational AI

3
Comments
4 min read
loading...