The objective is to apply a Gradient Tracking algorithm to achieve consensus optimization across multiple agents. This will demostrate how distributed systems can collaborate to determine a nonlinear classifier for data points in a feature space.
- Open file task_1.1_1.2
- Set the variable Task1_1 to 1
- You can set the following parameters as you wish from the file directly:
- NN (number of agents)
- alpha (learning rate)
- MAXITERS (Maximum Iterations for the update steps)
- graph_type (you can choose which type of graph you want among the agents from the following: "cycle", "path", "star", "complete")
- Now run the python file
- Open file task_1.1_1.2
- Set the variable Task_1_2 to 1
- You can set the following parameters as you wish from the file directly:
- mm_training (defines the size of the training dataset)
- alpha_w (learning rate for weights)
- alpha_b (learning rate for bias)
- MAXITERS (Maximum Iterations for the update steps)
- Now run the python file
- You can also choose the type of decision boundary in the dataset from the following:
- ellipse (This is an elliptical boundary, set it to 1 and curve to 0 to choose this)
- curve (This is a non-linear curved boundary, set it to 1 and ellipse to 0 to choose this)
- Open file task_1.3
- You can set the following parameters as you wish from the file directly:
- NN (number of agents)
- alpha (learning rate)
- MAXITERS (Maximum Iterations for the update steps)
- dataset_size (defines the size of the training dataset)
- graph_type (you can choose which type of graph you want among the agents from the following: "cycle", "path", "star", "complete")
- You can also choose the type of decision boundary in the dataset from the following:
- ellipse (This is an elliptical boundary, set it to 1 and curve to 0 to choose this)
- curve (This is a non-linear curved boundary, set it to 1 and ellipse to 0 to choose this)
- Execute the python file