Awni Hannun is a machine learning researcher and writer specializing in speech recognition, sequence models, and deep learning frameworks. He shares insights, tutorials, and analyses on ML concepts, algorithms, and practical implementations.
6 articles from this blog
An introduction to Fisher Information, a statistical concept that quantifies how much information data samples contain about unknown distribution parameters.
A forecast of speech recognition technology's evolution from 2010 to 2030, analyzing past progress and predicting future trends.
Explains the label bias problem in sequence models like MEMMs and how it affects predictions, leading to models like CRFs.
Practical tips for training sequence-to-sequence models with attention, focusing on debugging and ensuring the model learns to condition on input.
Argues that speech recognition hasn't reached human-level performance, highlighting persistent challenges with accents, noise, and semantic errors.
A comparison of PyTorch and TensorFlow deep learning frameworks, focusing on programmability, flexibility, and ease of use for different project scales.
