Mondrian Tree
A probabilistic data structure based on random splitting and local statistics, used for online classification and regression. It requires no fixed window or learning rate and adapts to concept drift.
Timeline (last 90 days)
A streaming learning engine based on a native JAX implementation was officially released, introducing a Mondrian tree routing mechanism.
The main source of the 50x speedup is rewriting the routing logic as pure functions traceable by JAX and using jit compilation and batched operations to achieve vectorized computation, rather than an algorithmic breakthrough.
The Mondrian tree was proposed by Balaji Lakshminarayanan and others around 2014, originally for online classification and regression.
During the construction of a Mondrian tree, the splitting dimension and splitting position are determined by uniform random numbers, so the tree structure itself carries randomness.
A Mondrian tree can grow naturally with the data stream without setting a window length or learning rate, and is more robust under concept drift than fixed-window methods.
All Facts (5)
The main source of the 50x speedup is rewriting the routing logic as pure functions traceable by JAX and using jit compilation and batched operations to achieve vectorized computation, rather than an algorithmic breakthrough.
50%UnverifiedThe Mondrian tree was proposed by Balaji Lakshminarayanan and others around 2014, originally for online classification and regression.
50%UnverifiedDuring the construction of a Mondrian tree, the splitting dimension and splitting position are determined by uniform random numbers, so the tree structure itself carries randomness.
50%UnverifiedA Mondrian tree can grow naturally with the data stream without setting a window length or learning rate, and is more robust under concept drift than fixed-window methods.
50%UnverifiedA streaming learning engine based on a native JAX implementation was officially released, introducing a Mondrian tree routing mechanism.
50%