concept task in category deep learning

appears as: task, tasks, task, tasks, The task
Deep Learning with Python, Second Edition

This is an excerpt from Manning's book Deep Learning with Python, Second Edition.

  • Input data points — For instance, if the task is speech recognition, these data points could be sound files of people speaking. If the task is image tagging, they could be pictures.
  • Two key concepts you’ll see frequently in future chapters are loss and optimizers. These are the two things you need to define before you begin feeding data into a model.

    • The loss is the quantity you’ll attempt to minimize during training, so it should represent a measure of success for the task you’re trying to solve.
    • The optimizer specifies the exact way in which the gradient of the loss will be used to update parameters: for instance, it could be the RMSProp optimizer, SGD with momentum, and so on.
  • Deep Learning with PyTorch

    This is an excerpt from Manning's book Deep Learning with PyTorch.

    The poorly defined term artificial intelligence covers a set of disciplines that have been subjected to a tremendous amount of research, scrutiny, confusion, fantastical hype, and sci-fi fearmongering. Reality is, of course, far more sanguine. It would be disingenuous to assert that today’s machines are learning to “think” in any human sense of the word. Rather, we’ve discovered a general class of algorithms that are able to approximate complicated, nonlinear processes very, very effectively, which we can use to automate tasks that were previously limited to humans.

    The key part is this: the cancers that we are trying to detect will always be nodules, either suspended in the very non-dense tissue of the lung or attached to the lung wall. That means we can limit our classifier to only nodules, rather than have it examine all tissue. Being able to restrict the scope of expected inputs will help our classifier learn the task at hand.

    Deep Learning with JavaScript: Neural networks in TensorFlow.js

    This is an excerpt from Manning's book Deep Learning with JavaScript: Neural networks in TensorFlow.js.

    Table 1.1. Examples of tasks in which accuracy improved significantly thanks to deep-learning techniques since the beginning of the deep-learning revolution around 2012. This list is by no means comprehensive. The pace of progress will undoubtedly continue in the coming months and years. (view table figure)

    Machine-learning task

    Representative deep-learning technology

    Where we use TensorFlow.js to perform a similar task in this book

    Categorizing the content of images Deep convolutional neural networks (convnets) such as ResNet[a] and Inception[b] reduced the error rate in the ImageNet classification task from ~25% in 2011 to below 5% in 2017.[c] Training convnets for MNIST (chapter 4); MobileNet inference and transfer learning (chapter 5)
    Localizing objects and images Variants of deep convnets[d] reduced localization error from 0.33 in 2012 to 0.06 in 2017. YOLO in TensorFlow.js (section 5.2)
    Translating one natural language to another Google’s neural machine translation (GNMT) reduced translation error by ~60% compared to the best traditional machine-translation techniques.[e] Long Short-Term Memory (LSTM)-based sequence-to-sequence models with attention mechanisms (chapter 9)
    Recognizing large-vocabulary, continuous speech An LSTM-based encoder-attention-decoder architecture achieves a lower word-error rate than the best non-deep-learning speech recognition system.[f] Attention-based LSTM small-vocabulary continuous speech recognition (chapter 9)
    Generating realistic-looking images Generative adversarial networks (GANs) are now capable of generating realistic-looking images based on training data (see https://github.com/junyanz/CycleGAN). Generating images using variational autoencoders (VAEs) and GANs (chapter 9)
    Generating music Recurrent neural networks (RNNs) and VAEs are helping create music scores and novel instrument sounds (see https://magenta.tensorflow.org/demos). Training LSTMs to generate text (chapter 9)
    Learning to play games Deep learning combined with reinforcement learning (RL) lets machines learn to play simple Atari games using raw pixels as the only input.[g] Combining deep learning and Monte Carlo tree search, Alpha-Zero reached a super-human level of Go purely through self-play.[h] Using RL to solve the cart-pole control problem and a snake video game (chapter 11)
    Diagnosing diseases using medical images Deep convnets were able to achieve specificity and sensitivity comparable to trained human ophthalmologists in diagnosing diabetic retinopathy based on images of patients’ retinas.[i] Transfer learning using a pretrained MobileNet image model (chapter 5).
  • AI is the study of automating cognitive tasks. Machine learning is a subfield of AI in which rules for performing a task such as image classification are discovered automatically by learning from examples in the training data.
  • A central problem in machine learning is how to transform the original representation of data into a representation more amenable to solving the task.
  • Grokking Deep Learning

    This is an excerpt from Manning's book Grokking Deep Learning.

    Given that deep learning is a subset of machine learning, what is machine learning? Most generally, it is what its name implies. Machine learning is a subfield of computer science wherein machines learn to perform tasks for which they were not explicitly programmed. In short, machines observe a pattern and attempt to imitate it in some way that can be either direct or indirect.

    What should you use the neural network to predict so that it learns more-interesting weight values for the word neurons? The task you’ll use to learn more-interesting weight values for the word neurons is a glorified fill-in-the blank task. Why use this? First, there’s nearly infinite training data (the internet), which means nearly infinite signal for the neural network to use to learn more-nuanced information about words. Furthermore, being able to accurately fill in the blank requires at least some notion of context about the real world.

    Deep Learning with R

    This is an excerpt from Manning's book Deep Learning with R.

  • Input data points—For instance, if the task is speech recognition, these data points could be sound files of people speaking. If the task is image tagging, they could be pictures.
  • Loss function (objective function)—The quantity that will be minimized during training. It represents a measure of success for the task at hand.