Terence Parr

Terence Parr Profile and History

Deep explanations of machine learning and related topics. Terence Parr is a tech lead at Google and until 2022 was a professor of data science / computer science at Univ. of San Francisco, where he was founding director of the MS in data science program in 2012. While he is best known for creating the ANTLR parser generator , Terence actually started out studying neural networks in grad school (19 87). After 30 years of parsing, he's back to machine learning and really enjoys trying to explain complex topics deeply and in the simplest possible way. Follow @the_antlr_guy . One of the biggest challenges when writing code to implement deep learning networks is getting all of the tensor (matrix and vector) dimensions to line up properly, even when using predefined network layers. This article describes a new library called TensorSensor that clarifies exceptions by augmenting messages and visualizing Python code to indicate the shape of tensor variables. It works with JAX, Tensorflow, PyTorch, and Numpy, as well as higher-level libraries like Keras and fastai. See also the TensorSensor implementation slides (PDF). Vanilla recurrent neural networks (RNNs) form the basis of more sophisticated models, such as LSTMs and GRUs. But, sometimes the neural network metaphor makes it less clear exactly what's going on. This articles explains RNNs without neural networks, stripping them down to its essencea series of vector transformations that result in embeddings for variable-length input vectors. I provide full PyTorch implementation notebooks that use just linear algebra and the autograd feature. Linear and logistic regression models are important because they are interpretable, fast, and form the basis of deep learning neural networks. Unfortunately, linear models have a tendency to chase outliers in the training data, which often leads to models that don't generalize well to new data. To produce models that generalize better, we all know to regularize our models. While there are lots of articles on the mechanics of ...