Learning in Brains and Machines (3): Synergistic and Modular Action 3

Learning in Brains and Machines (3): Synergistic and Modular Action
· Read in · 1796 words · All posts in series  ·  is a dance—precisely choreographed and executed—that we perform throughout our lives. This is the dance formed by our movements. Our movements are our actions and the final outcome of our decision making processes. Single actions are built into reusable sequences, sequences are composed into complex routines, routines are arranged into elegant choreographies, and so the complexity of ...

Learning in Brains and Machines (2): The Dogma of Sparsity 3

Learning in Brains and Machines (2): The Dogma of Sparsity
· Read in · 1700 words · collected posts · functioning of our brains, much like the intrigue of a political drama, is a neuronal house-of-cards. The halls of cognitive power are run by recurring alliances of neurons that deliberately conspire to control information processing and decision making. 'Suspicious coincidences' in neural activation—as the celebrated neuroscientist Horace Barlow observed—are abound; transparency in neural ...

Learning in Brains and Machines (1): Temporal Differences 12

Learning in Brains and Machines (1): Temporal Differences
· Read in · 1800 words · collected posts · We all make mistakes, and as is often said, only then can we learn. Our mistakes allow us to gain insight, and the ability to make better judgements and fewer mistakes in future. In their influential paper, the neuroscientists Robert Rescorla and Allan Wagner put this more succinctly, 'organisms only ...

A Year of Approximate Inference: Review of the NIPS 2015 Workshop 3

A Year of Approximate Inference: Review of the NIPS 2015 Workshop
inference lies no longer at the fringe. The importance of how we connect our observed data to the assumptions made by our statistical models—the task of inference—was a central part of this year's Neural Information Processing Systems (NIPS) conference: Zoubin Ghahramani's opening keynote set the stage and the pageant of inference included forecasting, compression, decision making, personalised modelling, and automating the ...

Machine Learning Trick of the Day (6): Tricks with Sticks 2

Machine Learning Trick of the Day (6): Tricks with Sticks
· Read in  · Our first encounters with probability are often through a collection of simple games. The games of probability are played with coins, dice, cards, balls and urns, and sticks and strings. Using these games, we built an intuition that allows us to reason and act in ways that account for randomness in the world. But ...

Machine Learning Trick of the Day (5): Log Derivative Trick 12

Machine Learning Trick of the Day (5): Log Derivative Trick
Machine learning involves manipulating probabilities. These probabilities are most often represented as normalised-probabilities or as log-probabilities. An ability to shrewdly alternate between these two representations is a vital step towards strengthening the probabilistic dexterity we need to solve modern machine learning problems. Today's trick, the log derivative trick, helps us to do just this, using the property of the derivative of the logarithm. This ...

Talk: Memory-based Bayesian Reasoning and Deep Learning

Talk: Memory-based Bayesian Reasoning and Deep Learning
A talk that explores the convergence of deep learning and Bayesian inference.  We'll take a statistical tour of deep learning, think about approximate Bayesian inference, and explore the idea of doing inference-with-memory and the different ways that this manifests itself in contemporary machine learning. Slides See the slides using this link. Abstract Deep learning and Bayesian ...

Machine Learning Trick of the Day (4): Reparameterisation Tricks 11

Machine Learning Trick of the Day (4): Reparameterisation Tricks
Our ability to rewrite statistical problems in an equivalent but different form, to reparameterise them, is one of the most general-purpose tools we have in mathematical statistics. We used reparameterisation in all the tricks we explored in this series so far: trick 1 re-expressed a log-partition function in terms of copies (replicas) of the marginal probability, trick 2 re-expressed a ...

Bayesian Reasoning and Deep Learning 4

Bayesian Reasoning and Deep Learning
I gave a talk entitled 'Bayesian Reasoning and Deep Learning' recently. Here is the abstract and the slides for interest. Slides Bayesian Reasoning and Deep Learning Abstract Deep learning and Bayesian machine learning are currently two of the most active areas of machine learning research. Deep learning provides a powerful class of models and an ...

Machine Learning Trick of the Day (3): Hutchinson's Trick 6

Machine Learning Trick of the Day (3): Hutchinson's Trick
Hutchinson's estimator  is a simple way to obtain a stochastic estimate of the trace of a matrix. This is a simple trick that uses randomisation to transform the algebraic problem of computing the trace into the statistical problem of computing an expectation of a quadratic function. The randomisation technique used in this trick is just one from a ...