Scribd Logo
Upload

Welcome to Scribd!

  • 11 views

    Assignment 3

    Uploaded by

    Seuneedhi Singh
    The document describes building a growing neural network tree using a dataset from Assignment 2. Small neural networks act as nodes that try to classify samples, with child nodes added to correct misclassified samples. The individual nodes are trained using a standard activation function. Growth is terminated to avoid overfitting using maximum depth and minimum accuracy thresholds. Results of the growing network, like loss and accuracy over training, validation, and test subsets, are plotted as the network grows.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content

    You might also like

    Assignment 3

    Uploaded by

    Seuneedhi Singh
    11 views3 pages
    The document describes building a growing neural network tree using a dataset from Assignment 2. Small neural networks act as nodes that try to classify samples, with child nodes added to correct misclassified samples. The individual nodes are trained using a standard activation function. Growth is terminated to avoid overfitting using maximum depth and minimum accuracy thresholds. Results of the growing network, like loss and accuracy over training, validation, and test subsets, are plotted as the network grows.

    Original Description:

    assignment

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    The document describes building a growing neural network tree using a dataset from Assignment 2. Small neural networks act as nodes that try to classify samples, with child nodes added to correct misclassified samples. The individual nodes are trained using a standard activation function. Growth is terminated to avoid overfitting using maximum depth and minimum accuracy thresholds. Results of the growing network, like loss and accuracy over training, validation, and test subsets, are plotted as the network grows.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as pdf or txt
    11 views3 pages

    Assignment 3

    Uploaded by

    Seuneedhi Singh
    The document describes building a growing neural network tree using a dataset from Assignment 2. Small neural networks act as nodes that try to classify samples, with child nodes added to correct misclassified samples. The individual nodes are trained using a standard activation function. Growth is terminated to avoid overfitting using maximum depth and minimum accuracy thresholds. Results of the growing network, like loss and accuracy over training, validation, and test subsets, are plotted as the network grows.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as pdf or txt
    of 3

    Assignment 3

    (by EEZ228477)

    Problem Statement: Consider your dataset that you built for Assignment 2. Use the same dataset
    to build a tree type growing neural network. At each node, learn a small neural network that tries to
    classify samples correctly, and child nodes are added to correct for misclassified samples. Explain
    how you will train the individual nodes while using a standard activation function, and then use the
    trained network as a node in a growing structure. Plot the loss function values and accuracies for
    training, validation, and test subsets, as the network is grown. Deduce a scheme to terminate growth
    to avoid overfitting. Please explain your approach and any assumptions in detail. Any missing detail
    may lead to hurdles in a fair evaluation. Use well designed figures to explain your approach and the
    growing network.

    Solution:

    (Link of the code at google Colab:


    https://colab.research.google.com/drive/1zdHZTDAIQoSy5sZ5JK9SOcYXvjgNTzRe?usp=shari
    ng )

    Figure of a neuron consisting of a conv layer and 2 dense layer


    Miss-classified data

    Miss-classified data

    The paragraph explains the formation of a small neural network which will be act as a node
    to build a growing tree like neural network. This small neuron unit consists of one
    convolutional layer and two dense layers. This is the minimum complexity architecture, we
    can consider. The results of a single node is that its training and validation accuracy is
    coming to be 98.29% and 98.36 % respectively.

    Now the formation of growing tree network is explained from here. The main part of the code
    is the grow tree function, which is recursive function responsible for building and growing
    neural network tree. The challenging part of growing neural network is splitting the
    data.Below are the two parameters which determines the growth of the neural network tree.

    max_depth: The maximum depth of the tree (the maximum number of child nodes a node
    can have). This prevents the tree from growing too deep.
    min_accuracy: The minimum acceptable accuracy on the validation set. If a node's validation
    accuracy drops below this threshold, we stop growing the tree further. This is a decision
    parameter which determines to keep a node or not.

    Fig: Results of the growing neural network

    You might also like