• FFT 1
• Fourier transform 1
• Galaxy Zoo 3
• Kaggle 4
• Lasagne 1
• MIR 2
• National Data Science Bowl 1
• Python 1
• Spotify 1
• Theano 9
• VAE 1
• arxiv 1
• audio 1
• autoencoder 1
• autoencoders 1
• autoregression 2
• code 1
• cold start problem 1
• collaborative filtering 1
• competition 4
• consistency models 1
• convnets 9
• convolution 4
• convolutional neural networks 9
• deep learning 22
• denoising 1
• diffusion 9
• diffusion model 1
• distillation 1
• fast Fourier transform 1
• generative models 12
• geometry 1
• guidance 2
• internship 1
• language 1
• latent space 1
• likelihood 2
• music 2
• music information retrieval 2
• music recommendation 1
• natural images 1
• neural networks 1
• noise schedule 1
• optimization 1
• paper 1
• parameterisation 1
• plankton 1
• probability 2
• raw audio 1
• score function 5
• slides 1
• spectral analysis 1
• spectrum 1
• talk 1
• typicality 2
• vectors 1
• waveform 1

FFT

• Even faster convolutions in Theano using FFTs

Fourier transform

• Diffusion is spectral autoregression

Galaxy Zoo

• Paper about my Galaxy Challenge solution
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge

Kaggle

• Paper about my Galaxy Challenge solution
• Classifying plankton with deep neural networks
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge

Lasagne

• New Lasagne feature: arbitrary expressions as layer parameters

MIR

• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup

National Data Science Bowl

• Classifying plankton with deep neural networks

Python

• Galaxy Zoo Challenge: code published

Spotify

• Recommending music on Spotify with deep learning

Theano

• New Lasagne feature: arbitrary expressions as layer parameters
• Classifying plankton with deep neural networks
• The fastest convolutions in Theano with meta-optimization
• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup
• Even faster convolutions in Theano using FFTs
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge
• 3x faster convolutions in Theano

VAE

• Generative modelling in latent space

arxiv

• Paper about my Galaxy Challenge solution

audio

• Generating music in the waveform domain

autoencoder

• Generative modelling in latent space

autoencoders

• Diffusion models are autoencoders

autoregression

• Generative modelling in latent space
• Diffusion is spectral autoregression

code

• Galaxy Zoo Challenge: code published

cold start problem

• Recommending music on Spotify with deep learning

collaborative filtering

• Recommending music on Spotify with deep learning

competition

• Paper about my Galaxy Challenge solution
• Classifying plankton with deep neural networks
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge

consistency models

• The paradox of diffusion distillation

convnets

• Paper about my Galaxy Challenge solution
• Classifying plankton with deep neural networks
• The fastest convolutions in Theano with meta-optimization
• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup
• Even faster convolutions in Theano using FFTs
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge
• 3x faster convolutions in Theano

convolution

• The fastest convolutions in Theano with meta-optimization
• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup
• Even faster convolutions in Theano using FFTs

convolutional neural networks

• Paper about my Galaxy Challenge solution
• Classifying plankton with deep neural networks
• The fastest convolutions in Theano with meta-optimization
• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup
• Even faster convolutions in Theano using FFTs
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge
• 3x faster convolutions in Theano

deep learning

• Generative modelling in latent space
• Diffusion is spectral autoregression
• Noise schedules considered harmful
• The paradox of diffusion distillation
• The geometry of diffusion guidance
• Perspectives on diffusion
• Diffusion language models
• Guidance: a cheat code for diffusion models
• Diffusion models are autoencoders
• Musings on typicality
• Addendum: quantifying our flawed intuitions
• Generating music in the waveform domain
• New Lasagne feature: arbitrary expressions as layer parameters
• Paper about my Galaxy Challenge solution
• Classifying plankton with deep neural networks
• The fastest convolutions in Theano with meta-optimization
• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup
• Even faster convolutions in Theano using FFTs
• Galaxy Zoo Challenge: code published
• My solution for the Galaxy Zoo challenge
• 3x faster convolutions in Theano

denoising

• Diffusion models are autoencoders

diffusion

• Generative modelling in latent space
• Diffusion is spectral autoregression
• Noise schedules considered harmful
• The paradox of diffusion distillation
• The geometry of diffusion guidance
• Perspectives on diffusion
• Diffusion language models
• Guidance: a cheat code for diffusion models
• Diffusion models are autoencoders

diffusion model

• Noise schedules considered harmful

distillation

• The paradox of diffusion distillation

fast Fourier transform

• Even faster convolutions in Theano using FFTs

generative models

• Generative modelling in latent space
• Diffusion is spectral autoregression
• Noise schedules considered harmful
• The paradox of diffusion distillation
• The geometry of diffusion guidance
• Perspectives on diffusion
• Diffusion language models
• Guidance: a cheat code for diffusion models
• Diffusion models are autoencoders
• Musings on typicality
• Addendum: quantifying our flawed intuitions
• Generating music in the waveform domain

geometry

• The geometry of diffusion guidance

guidance

• The geometry of diffusion guidance
• Guidance: a cheat code for diffusion models

internship

• Recommending music on Spotify with deep learning

language

• Diffusion language models

latent space

• Generative modelling in latent space

likelihood

• Musings on typicality
• Addendum: quantifying our flawed intuitions

music

• Generating music in the waveform domain
• Recommending music on Spotify with deep learning

music information retrieval

• Recommending music on Spotify with deep learning
• Slides for my talk at the Deep Learning London meetup

music recommendation

• Recommending music on Spotify with deep learning

natural images

• Diffusion is spectral autoregression

neural networks

• New Lasagne feature: arbitrary expressions as layer parameters

noise schedule

• Noise schedules considered harmful

optimization

• The fastest convolutions in Theano with meta-optimization

paper

• Paper about my Galaxy Challenge solution

parameterisation

• Noise schedules considered harmful

plankton

• Classifying plankton with deep neural networks

probability

• Musings on typicality
• Addendum: quantifying our flawed intuitions

raw audio

• Generating music in the waveform domain

score function

• The geometry of diffusion guidance
• Perspectives on diffusion
• Diffusion language models
• Guidance: a cheat code for diffusion models
• Diffusion models are autoencoders

slides

• Slides for my talk at the Deep Learning London meetup

spectral analysis

• Diffusion is spectral autoregression

spectrum

• Diffusion is spectral autoregression

talk

• Slides for my talk at the Deep Learning London meetup

typicality

• Musings on typicality
• Addendum: quantifying our flawed intuitions

vectors

• The geometry of diffusion guidance

waveform

• Generating music in the waveform domain