This is a beginner-friendly PyTorch project that demonstrates how a simple neural network learns using linear layers, loss functions, and optimizers. It walks through the core concepts of machine learning from the ground up, using a hands-on example where a model learns to predict a value.

I created this project as a personal learning exercise to understand the fundamentals of how machine learning works under the hood, especially in PyTorch. I focused on:

• Understanding how loss functions measure performance
• How optimizers adjust the model to reduce error
• How training across multiple epochs improves accuracy
• Writing and understanding a training loop

• A simple linear model is created using torch.nn.Linear
• Input and target values are defined
• The model makes predictions
• Mean Squared Error (nn.MSELoss) is used to calculate how wrong the predictions are
• An optimizer (torch.optim.SGD) updates the model based on the error
• A training loop repeats this process across multiple epochs

• Python 3.x
• PyTorch
• (Optional) matplotlib for visualizing loss

Install dependencies using:

pip install torch matplotlib

Clone the repo and run the file:

python simple_pytorch_model.py

Epoch 1, Loss: 11.3393 Epoch 2, Loss: 8.7214 Epoch 3, Loss: 6.7013 ... Epoch 10, Loss: 1.7020

📈 Concepts Covered

-Tensors in PyTorch -Linear regression -Loss functions (MSE) -Optimizers (SGD) -Gradient descent -Backpropagation -Epochs and model training

• Add multiple training examples

• Support for batch training

• Visualize the loss with matplotlib

• Predict values in a time series (e.g., sine wave)

• Wrap code into classes for reusability

• Turn model into full time series model with dataset