AdaCat: Adaptive categorical discretization for autoregressive models
Autoregressive generative models can estimate complex continuous data distributions, like
trajectory rollouts in an RL environment, image intensities, and audio. Most state-of-the-art
models discretize continuous data into several bins and use categorical distributions over
the bins to approximate the continuous data distribution. The advantage is that the
categorical distribution can easily express multiple modes and are straightforward to
optimize. However, such approximation cannot express sharp changes in density without …
trajectory rollouts in an RL environment, image intensities, and audio. Most state-of-the-art
models discretize continuous data into several bins and use categorical distributions over
the bins to approximate the continuous data distribution. The advantage is that the
categorical distribution can easily express multiple modes and are straightforward to
optimize. However, such approximation cannot express sharp changes in density without …
[PDF][PDF] web AdaCat: Adaptive Categorical Discretization for Autoregressive Models
Q Li, A Jain, P Abbeel - ajayj.com
… • Autoregressive generative models can estimate complex data distributions such as
language, audio and images. • Most existing methods either operate on discrete data
distribution or discretize continuous data into several bins and use categorical distributions
over the bins to approximate the continuous data distribution. … AdaCat is a mixture of k non-overlapping
truncated uniforms (w, h ∈ R k ) …
language, audio and images. • Most existing methods either operate on discrete data
distribution or discretize continuous data into several bins and use categorical distributions
over the bins to approximate the continuous data distribution. … AdaCat is a mixture of k non-overlapping
truncated uniforms (w, h ∈ R k ) …
Showing the best results for this search. See all results