Jump to: Overview Evolving a hidden state over time Common structures of recurrent networks Bidirectionality Limitations Further reading Overview Previously, I've written about feed-forward neural networks as a generic function approximator and convolutional neural networks for efficiently extracting local information from data. In this
Jump to: What is nearest neighbors search? K-d trees Quantization Product quantization Handling multi-modal data Locally optimized product quantization Common datasets Further reading What is nearest neighbors search? In the world of deep learning, we often use neural networks to learn representations of objects
When evaluating a standard machine learning model, we usually classify our predictions into four categories: true positives, false positives, true negatives, and false negatives. However, for the dense prediction task of image segmentation, it's not immediately clear what counts as a "true positive&
In this post, I'll discuss commonly used architectures for convolutional networks. As you'll see, almost all CNN architectures follow the same general design principles of successively applying convolutional layers to the input, periodically downsampling the spatial dimensions while increasing the number of feature maps.
In my introductory post on autoencoders, I discussed various models (undercomplete, sparse, denoising, contractive) which take data as input and discover some latent state representation of that data. More specifically, our input data is converted into an encoding vector where each dimension represents some
Autoencoders are an unsupervised learning technique in which we leverage neural networks for the task of representation learning. Specifically, we'll design a neural network architecture such that we impose a bottleneck in the network which forces a compressed knowledge representation of the original input.
In the previous post, I discussed two different learning methods for reinforcement learning, Monte Carlo learning and temporal difference learning. I then provided a unifying view by considering $n$-step TD learning and establishing hybrid learning method, $TD\left( \lambda \right)$. These methods are