Deep Learning Lab 2: Autograd - automatic differentiation
Deep Learning Lab 2
This lab covers pyTorch automatic differentiation (autograd). At the end a simple non-linear regression model is presented. In details the following topics are covered:
- Visualizing the computational graph
- Gradient accumulation
- Non-linear regression with a neural network
Non-linear regression
The sample in the Jupyter notebook present the non-linear model below:
\(y = -(0.1 x_1^2 + 3 sin(0.1 x_2) +0.1)\)
We will solve this problem using a simple neural network using a single hidden layer with no activation function:
Here is the link to access the notebook on Google colab. It doesn’t work directly on github.