Scribd Logo
Upload

Welcome to Scribd!

  • 112 views

    Syllabus - Deep Learning and Edge Intelligence

    Uploaded by

    Nilesh Nagrale
    The document outlines a course on Deep Learning and Edge Intelligence that covers artificial neural networks, convolutional neural networks, recurrent neural networks, and deploying models to edge devices. The course aims to teach students how to design and implement deep learning models for applications like classification, regression, and computer vision, and how to optimize models for deployment on edge platforms with considerations for accuracy, size, and hardware constraints. Practical work involves implementing deep learning architectures and developing applications using edge devices.

    Copyright:

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

    Syllabus - Deep Learning and Edge Intelligence

    Uploaded by

    Nilesh Nagrale
    112 views3 pages
    The document outlines a course on Deep Learning and Edge Intelligence that covers artificial neural networks, convolutional neural networks, recurrent neural networks, and deploying models to edge devices. The course aims to teach students how to design and implement deep learning models for applications like classification, regression, and computer vision, and how to optimize models for deployment on edge platforms with considerations for accuracy, size, and hardware constraints. Practical work involves implementing deep learning architectures and developing applications using edge devices.

    Original Title

    Syllabus_Deep Learning and Edge Intelligence (2)

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document outlines a course on Deep Learning and Edge Intelligence that covers artificial neural networks, convolutional neural networks, recurrent neural networks, and deploying models to edge devices. The course aims to teach students how to design and implement deep learning models for applications like classification, regression, and computer vision, and how to optimize models for deployment on edge platforms with considerations for accuracy, size, and hardware constraints. Practical work involves implementing deep learning architectures and developing applications using edge devices.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    112 views3 pages

    Syllabus - Deep Learning and Edge Intelligence

    Uploaded by

    Nilesh Nagrale
    The document outlines a course on Deep Learning and Edge Intelligence that covers artificial neural networks, convolutional neural networks, recurrent neural networks, and deploying models to edge devices. The course aims to teach students how to design and implement deep learning models for applications like classification, regression, and computer vision, and how to optimize models for deployment on edge platforms with considerations for accuracy, size, and hardware constraints. Practical work involves implementing deep learning architectures and developing applications using edge devices.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 3

    [ET-] Deep Learning and Edge Intelligence

    Teaching Scheme Examination Scheme


    Lectures: 2 hrs./week Test I - 20 Marks
    Test II - 20 Marks
    End Sem Exam - 60 marks

    Course Outcomes:
    At the end of the course, students will demonstrate the ability to
    1. Design and implement FCNN models for applications such as classification and
    regression using ML/DL platform.
    2. Apply regularization techniques for improving the model performance on
    validation/test datasets.
    3. Design and/or select CNNs for various computer vision applications.
    4. Apply and analyze RNNs for various applications that involve temporal/sequence
    learning tasks.5
    5. Deploy ML/DL models designed for vision and sequence learning applications on edge
    platforms.

    Unit 1 (4 hrs)
    Artificial Neural Networks (ANN)
    Introduction, Perceptron Training Rule, Sigmoid neuron, fully connected neural network,
    Activation functions, Loss functions for regression and classification, Gradient Descent and
    Backpropagation, Batch and mini-batch processing, Stochastic Gradient Descent

    Unit 2 (4 hrs)
    Regularization Techniques
    Optimization and Regularization: Overfitting, Cross-Validation, Feature Selection,
    Regularization (L2 penalty, dropout, ensembles, data augmentation techniques)
    Hyperparameter tunning

    Unit 3 (4 hrs)
    Convolutional Neural Networks (CNN)
    Convolutional Neural Network, Weight Initialization, Visualizing filters of a CNN, Guided
    Backpropagation, Autoencoders, Regularization in autoencoders, Denoising Autoencoders,
    Standard CNN models such as Alexnet, VGG net, Inception net, Resnet.

    Unit 4 (4 hrs)
    Recurrent neural networks (RNN)
    Recurrent neural networks, sequence modelling, backpropagation through time;
    Hyperparameters, Vanishing/exploding gradient problem, Gradient clipping, Long-short term
    memory (LSTM)

    Unit 5 (6 hrs)
    Edge Devices and Networking
    Edge computing, Advantages, Technologies involved in Edge development, Internet web and
    Networking technologies, Infrastructure, Role of M2M in Edge Intelligence, Overview of
    supported Hardware like NVIDIA Jetson Nano, TX2, Xavier NX Developer boards
    Unit 6 (6 hrs)
    Edge AI Solution
    Model design with a focus on a reduced number of parameters, Efficient DNNs such as
    Squeeze net, Neural architecture search (NAS), Model compression: Lowering precision and
    fewer weights (pruning), post-training quantization tools e.g., TensorFlow Lite, Device
    consideration for deploying ML/DL algorithms, Ways of balancing model architecture
    (accuracy, size, operation type) requirements with device programmability, throughput,
    power consumption, and cost.

    Textbooks:
     Bishop Christopher, “Neural Networks for Pattern Recognition”, New York, NY: Oxford
    University, Press, ISBN: 9780198538646

    Reference books:
     Deep Learning by Ian Goodfellow, Yoshua Bengio and Aaron Courville published by
    MIT Press, 2016, ISBN: 9780262035613
     Raj Kamal, “Internet of Things: Architecture and Design”, McGraw Hill
     Online references for edge computing and intelligence
     Online references at www.tensorflow.org, www.pytorch.org
    (ET-) Deep Learning and Edge Intelligence Lab
    Teaching Scheme Examination Scheme
    Practical: 2 hrs./week Term work - 50 Marks
    Practical/Oral – 50 marks
    Course Outcomes:
    At the end of the laboratory work, students will demonstrate the ability to
    1. Design and interpret the functioning of Deep Learning models, namely with CNN &
    RNN
    2. Develop Classification and Regression applications to categorize and predict on test
    data samples.
    3. Design and develop various AI applications using NVIDIA Jetson Nano /Raspberry-Pi
    and their various sensor interfacing
    4. Analyze and evaluate the deployment of edge devices in the modern AI world.

    List of Experiments:
    1. Implement the gradient descent algorithm for a three-stage fully connected neural
    network using python and basic packages such as NumPy and matplotlib.
    2. Implement Deep Neural Network for classification and regression tasks such as
    complex human activity recognition (HAR), temperature forecasting, etc.
    3. Apply and Implement Deep Convolutional Neural Network (DCNN) architectures for
    object detection and classification.
    4. Implement twitter spam detection using natural language processing by
    Autoencoder model.
    5. Implement U-NET architecture for robust lungs segmentation in chest X-rays.
    6. Implement anomaly activity detection through videos using RNNs and their variants.
    7. Apply and Implement handwritten Text Recognition with TensorFlow using Deep
    Learning.
    Project List: Choose any One
    8. Develop an Application on NVIDIA Jetson Nano /Raspberry-Pi to capture the values
    of temperature sensor after every 15 sec of time interval, store these values in .csv
    format and predict the temperature at particular time t using linear regression
    analysis.
    9. Develop an Application on NVIDIA Jetson Nano /Raspberry-Pi for Object Recognition
    based smart AI BoT
    10. Develop an Application on NVIDIA Jetson Nano /Raspberry-Pi for Smart Agriculture

    You might also like