Skip to content

README.md

Latest commit

 

History

History
39 lines (35 loc) · 2.17 KB

README.md

File metadata and controls

39 lines (35 loc) · 2.17 KB

Distributed-Classification-via-Logistic-Regression

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.

Task 1.1: Gradient Tracking Algorithm

  1. Open file task_1.1_1.2
  2. Set the variable Task1_1 to 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)
    • graph_type (you can choose which type of graph you want among the agents from the following: "cycle", "path", "star", "complete")
  4. Now run the python file

Task 1.2: Centralized Gradient Method using the logistic regression cost function

  1. Open file task_1.1_1.2
  2. Set the variable Task_1_2 to 1
  3. 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
  4. 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)

Task 1.3: Gradient Tracking Method on the Logistic Regression Cost Function

  1. Open file task_1.3
  2. 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")
  3. 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)
  4. Execute the python file