Scribd Logo
Upload

Welcome to Scribd!

  • 5 views

    Deep Learning Paper Review

    Uploaded by

    Jonnah Mann Dèò
    This paper discusses deep learning as a part of machine learning. It explains how deep learning uses multiple hidden layers to classify features in data and the different types of neural networks used for tasks like image recognition. The paper focuses on supervised learning methods like stochastic gradient descent and explores challenges like training recurrent neural networks.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Deep Learning Paper Review

    Uploaded by

    Jonnah Mann Dèò
    5 views1 page
    This paper discusses deep learning as a part of machine learning. It explains how deep learning uses multiple hidden layers to classify features in data and the different types of neural networks used for tasks like image recognition. The paper focuses on supervised learning methods like stochastic gradient descent and explores challenges like training recurrent neural networks.

    Original Title

    DEEP LEARNING PAPER REVIEW

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This paper discusses deep learning as a part of machine learning. It explains how deep learning uses multiple hidden layers to classify features in data and the different types of neural networks used for tasks like image recognition. The paper focuses on supervised learning methods like stochastic gradient descent and explores challenges like training recurrent neural networks.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    5 views1 page

    Deep Learning Paper Review

    Uploaded by

    Jonnah Mann Dèò
    This paper discusses deep learning as a part of machine learning. It explains how deep learning uses multiple hidden layers to classify features in data and the different types of neural networks used for tasks like image recognition. The paper focuses on supervised learning methods like stochastic gradient descent and explores challenges like training recurrent neural networks.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 1

    DEEP LEARNING PAPER REVIEW

    This paper is based on deep learning a constituent of machine learning, to understand


    deep learning one must understand how machine learning works and more so, how the two are
    connected. Technology has changed over the years and automation has become the norm, and
    machine learning has been incorporated into web searches, e-commerce websites all with the
    help of deep learning. The aspect of machine learning is not necessarily based on the engineer's
    design of the system but learning is based on the raw data provided, deep learning provides
    multiple layers that are responsible for detecting the key aspects used in the classification of
    various features available in the data, there are three major types of layers namely; input, output
    and lastly hidden layers whose number is determined by the type of data that may be in the field
    of speech or image recognition.
    Machine learning comes in different forms in this paper supervised learning was the main
    focus, just like the name suggests the model is exposed to data with what it needs to learn and
    uses that information to predict if exposed to similar data. One of the methods used is the
    stochastic gradient descent which computes the outputs and errors and finds the average gradient
    and it adjusts itself accordingly this method tends to find the minimum gradient at the least time
    possible. A linear classifier may also be used but it requires a good feature extractor to solve
    selective- variance dilemma this may not happen for SDG.
    Multilayer data also provided a challenge in the field of deep learning this was met head-
    on by using methods like Backpropagation which helps reach local minimal and also that was
    Convolutional neural networks which is a method crafted from the field of visual neuroscience
    where it used layers and uses pooling method to constitute. Convolutional neural networks are
    good in the application of image and motion processing used in robots, for Recurrent neural
    networks, the method is used to detect speech and language since it processes input sequences
    one at a time but they are problematic when it comes to training them. Long short-term memory
    (LSTM) networks usually solve the issue by providing the memory needed to remember inputs.
    The paper creates a great picture and elaborates more on deep learning but it tends to rely
    more on a supervised approach which in this case doesn't cover all the fields of machine
    learning. The future of machine learning will get broader as time goes by given the fact that
    automation of machines from cars, industrial applications, instruments of learning, and looking
    from the dark side of war but all in all deep learning will keep on improving, and as this paper
    has shown the ground is set for future generation to exploit and grow the field into greater
    heights.

    You might also like