Exploring Intelligent Systems Through Simulation

Neural Networks: A Foundation

At the heart of many AI systems lie artificial neural networks. These models are inspired by the structure of biological brains, consisting of interconnected nodes (neurons) organized in layers.

Our simulation allows you to experiment with different network architectures – from simple perceptrons to multi-layered deep neural networks. You can adjust parameters like learning rate and momentum to observe their impact on model performance.

f(x) = Σ(wᵢ * xᵢ) + b  (Perceptron Activation Function)

Training Algorithms: Gradient Descent

Machine learning models learn through training, adjusting their internal parameters to minimize error. A common technique is gradient descent.

The simulation demonstrates how gradient descent iteratively updates the weights of a neural network based on the difference between predicted and actual outputs. This process aims to find the optimal set of weights that minimizes the loss function.

Δw = -η * ∇L(w)  (Gradient Descent Update Rule)

Decision Trees: Rule-Based Learning

Decision trees provide another approach to machine learning, creating a series of rules based on input features to predict outcomes.

The simulation lets you build decision trees by splitting data based on feature values. You can visualize the tree structure and evaluate its accuracy using metrics like Gini impurity or information gain.

Entropy = - Σ pᵢ * log₂(pᵢ)  (Measure of Impurity)

Model Evaluation & Tuning

Once a model is trained, it’s crucial to evaluate its performance and potentially tune its parameters for optimal results.

The simulation provides tools to assess accuracy, precision, recall, and F1-score. Experiment with different training datasets and hyperparameters to see how they affect the model's ability to generalize to unseen data.

Frequently asked questions

Try it live

Everything above runs in your browser — open SPH Fluid and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.