Udacity has announced a new course every day for the past two weeks. Each course will give you job-ready skills in less than a month and offers a 50% discount for a limited time.
The premise behind the 14 courses launched by Udacity is that you can learn practical technical skills in less than a month. It covers a wide range of topics including data science, AI, programming, cybersecurity and more.
All three of my picks today are AI-related. So, if you want to know about the other 11, here’s the complete list of Udacity short-term online courses that have enrollment dates of his June 7th and are being introduced at half the regular price. $200 instead of $399 per month.
I make no apologies for choosing an AI-related course. Some are wary of the pace of development towards an efficient and effective artificial intelligence, but we are embracing artificial intelligence, how it works, and how we can use it to our advantage, Understanding what you can get under control seems like a more reasonable approach. For some, it’s an even better idea to go inside and seek out his career on the AI frontier.
Udacity courses are created in collaboration with technology companies and taught by expert instructors. Learners receive support from technical mentors and individualized project feedback with line-by-line code reviews. Udacity claims:
Our learning-by-doing approach incorporates projects based on real-world scenarios. These are open-ended and require critical thinking, rather than just multiple-choice questions scored by a bot. Just like a workplace.
These new courses are designed to be completed in less than a month, but are still project-based and highly hands-on. Course projects are listed first in the overview below, followed by lessons that impart the necessary knowledge.
Construction of generative adversarial networks
At an advanced level, this course teaches how to build and train various generative adversarial network architectures to generate new images. Prerequisites include knowledge of convolutional neural networks (CNNs), intermediate Python proficiency, and experience with PyTorch.
Course Project: Face Generation
Leverage various skills learned during the course to build and train a custom GAN architecture on the CelebA dataset. Build a custom GAN architecture containing generators and discriminators. Experiment with different loss functions discovered during the course, including binary cross-entropy loss and Wasserstein loss. Finally, we take advantage of some methods we’ve learned to stabilize the training, such as label smoothing.
-
Generative Adversarial Network
Build generators and discriminators using fully connected layers. Implement a loss function and train a custom GAN on the MNIST dataset.
-
Training a Deep Convolutional GAN
Construct generators and discriminators using convolution, batch normalization, and fully connected layers. We will train a DCGAN model on the CIFAR10 dataset, implement an evaluation metric, and evaluate the generated samples.
-
image to image conversion
Implement an unpaired data loader. Construct a CycleGAN generator using residual connections and an encoder/decoder structure. We train a CycleGAN model on the Summer2Winter Yosemite dataset.
-
Latest GANs
We implement Wasserstein loss and gradient penalties and build a ProGAN generator. Implement a StyleGAN component (adaptive instance normalization).
Deep learning overview
At an intermediate level, this course delves deep into fundamental theoretical and practical topics related to deep learning. Prerequisites include intermediate Python skills and experience with Jupyter notebooks, Pandas, NumPy, and machine learning frameworks.
Course Project: Developing a Handwritten Digit Classifier Using PyTorch
Develop a handwritten digit recognition system with PyTorch. Then use your data preprocessing skills to properly load the data for use in your model. Develop a neural network using PyTorch and create a training loop to train the model on the loaded data. Finally, we apply advanced training techniques to improve the accuracy of the test set.
-
deep learning
Explain the difference between artificial intelligence, machine learning, and deep learning. Recognize the power of deep learning by reviewing popular examples of deep learning applications.
-
Minimizing the error function with gradient descent
We use PyTorch to preprocess the data and measure model performance using maximum likelihood, cross-entropy, and probability. Apply gradient descent to minimize the error. Implement the backpropagation algorithm and identify the key components of the perceptron.
-
Neural network overview
Describe important neural network concepts and design neural network architectures. Differentiate the problem based on the purpose of the model.Implement an architecture that suits the model’s purpose
-
Neural network training
Define a loss function and optimization method for training a neural network. Distinguish between overfitting and underfitting and identify the cause of each. Optimize the training process using early stopping, regularization, dropout, learning rate decay, and momentum. Distinguish between batch gradient descent and stochastic gradient descent, build a neural network in PyTorch and run the data on it. Test and validate the trained network to make sure it generalizes.
Advanced computer vision and deep learning
In this course, you will learn how to combine CNN and RNN networks to build an automatic image captioning application. Learners should have knowledge of Probability Fundamentals, Deep Learning Frameworks, Intermediate Her Python, Neural Network Fundamentals, and Object Oriented Programming Fundamentals.
Course Project: Image Captioning
We combine CNN and RNN knowledge to build a deep learning model that generates captions from input images. Creating image captions requires a learner to create a complex deep learning model containing her two components. One is a CNN that transforms an input image into a set of features, and the other is an RNN that transforms those features into a rich descriptive language. This project implements these state-of-the-art deep learning architectures.
-
Advanced CNN architecture
Describe the progress of CNN architecture. Understand region-based CNN, Fast R-CNN, and Faster-R-CNN that enable fast local object recognition in images.
-
Yolo
Understand grids, sliding windows, and bounding boxes in object detection. Implements YOLO, a real-time object detection algorithm.
-
of RNNs
Understand how recurrent neural networks learn from ordered data sequences. Identify RNN applications in deep learning. Understand how feedforward and backpropagation over time works, as well as the RNN unfolding model.
-
long short-term memory network (LSTM)
Explore ways to incorporate memory into deep learning models and implement long- and short-term memory networks.
-
Hyperparameter
Update important hyperparameters such as learning rate, epochs, layers, and understand RNN hyperparameters.
-
Optional: Attention Mechanism
Understand how attention allows the model to focus on specific parts of the input data. Learn where attention helps in natural language and computer vision applications. Describes Attention and its encoders and decoders that power applications such as text translation.Understand basic attention methods such as additive attention, badanau and luong attention
-
image caption
Understand image captions and tokenize captions and words. Combine CNNs and RNNs to build complex caption models.
If you haven’t yet met all the prerequisites required for these courses, our recently introduced chatbot, Udacity GPT, will provide recommendations.
For more information
Udacity Short-Term Online Course
Construction of generative adversarial networks
