Learning Overcomplete HMMs

11/07/2017
by   Vatsal Sharan, et al.
0

We study the problem of learning overcomplete HMMs---those that have many hidden states but a small output alphabet. Despite having significant practical importance, such HMMs are poorly understood with no known positive or negative results for efficient learning. In this paper, we present several new results---both positive and negative---which help define the boundaries between the tractable and intractable settings. Specifically, we show positive results for a large subclass of HMMs whose transition matrices are sparse, well-conditioned, and have small probability mass on short cycles. On the other hand, we show that learning is impossible given only a polynomial number of samples for HMMs with a small output alphabet and whose transition matrices are random regular graphs with large degree. We also discuss these results in the context of learning HMMs which can capture long-term dependencies.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
01/21/2022

Positive and negative cycles in Boolean networks

We review and discuss some results about the influence of positive and n...
research
02/11/2019

Election Manipulation on Social Networks with Positive and Negative Messages

We study the problem of election control through social influence when t...
research
03/05/2019

Reducing the domination number of graphs via edge contractions

In this paper, we study the following problem: given a connected graph G...
research
01/03/2017

Clustering Signed Networks with the Geometric Mean of Laplacians

Signed networks allow to model positive and negative relationships. We a...
research
11/05/2021

Geometric construction of canonical 3D gadgets in origami extrusions

In a series of our three previous papers, we presented several construct...
research
03/10/2022

Synchronizing Boolean networks asynchronously

The asynchronous automaton associated with a Boolean network f:{0,1}^n→{...
research
02/22/2022

Exact Community Recovery over Signed Graphs

Signed graphs encode similarity and dissimilarity relationships among di...

Please sign up or login with your details

Forgot password? Click here to reset