A scalable bayesian inference accelerator for unsupervised learning
2020 IEEE Hot Chips 32 Symposium (HCS), 2020•computer.org
… Machine learning allows us to extract useful information from data without much human
intervention.1Did we solve the problem?Ref: CIFAR-10Up to 80% of time of a data scientist is
spent on sourcing and cleaning dataDeep … Representing and computing with
uncertaintyAutonomous DrivingApplications •Biomedical •Robotics •Autonomous driving •FinanceProblem
CPUs and GPUs are inefficient for Bayesian inference5PGMA: Probabilistic Graphical Models
Accelerator•First silicon accelerator for Bayesian inference •Algorithm-hardware co-design for …
intervention.1Did we solve the problem?Ref: CIFAR-10Up to 80% of time of a data scientist is
spent on sourcing and cleaning dataDeep … Representing and computing with
uncertaintyAutonomous DrivingApplications •Biomedical •Robotics •Autonomous driving •FinanceProblem
CPUs and GPUs are inefficient for Bayesian inference5PGMA: Probabilistic Graphical Models
Accelerator•First silicon accelerator for Bayesian inference •Algorithm-hardware co-design for …
Gibbs sampling on MRF 1: Initialize x 0 2: for t= 0 to T do 3: for i= 0 to N do 4: x (t+ 1) i
computer.org
Showing the best result for this search. See all results