	US 12,293,308 B2
	Bit prediction method and system using a statistical model
Georgiy Buyanovskiy, Palo Alto, CA (US)
Assigned to FOVIA, INC., Palo Alto, CA (US)
Filed by FOVIA, INC., Palo Alto, CA (US)
Filed on Oct. 5, 2018, as Appl. No. 16/153,382.
Claims priority of provisional application 62/569,894, filed on Oct. 9, 2017.
Prior Publication US 2019/0108455 A1, Apr. 11, 2019
Int. Cl. G06N 7/01 (2023.01); G06F 16/22 (2019.01); G06F 16/31 (2019.01); H03M 7/30 (2006.01); G06F 30/20 (2020.01)

CPC G06N 7/01 (2023.01) [G06F 16/2246 (2019.01); G06F 16/31 (2019.01); H03M 7/3068 (2013.01); G06F 30/20 (2020.01)]

14 Claims

1. A method for data compression using providing a representation of a trie-structure, comprising:

at an electronic device having one or more processors:

receiving a data bit of a bit stream for a data compression optimization procedure using a representation of a trie-structure;

determining a probability of a binary event for each of a plurality of trie-nodes of a trie-structure based on a composition of probabilities of each context of a plurality of contexts of trie, wherein the composition includes a weighted sum of the probabilities based on weighting coefficients;

entropy coding the data bit based on a composite probability in order to tune weighting coefficients to improve compression and a prediction accuracy of a next bit;

adjusting and updating the probability of the binary event for each of the plurality of trie-nodes of the trie-structure, wherein:

adjusting the probability of each context of the plurality of contexts of trie is based on a respective weighted composite probability and a respective adjustment coefficient,

updating the probability of each context of the plurality of contexts of trie is based on the adjusted probability of each context, the probability of the binary event, and a respective update coefficient, and

updating the probability of each trie-node of the plurality of trie-nodes is based on a trie-node probability, the probability of the binary event, and a respective update coefficient; and

determining whether a current trie-node of the trie-structure has a reference to a first context trie-node corresponding to a value of the data bit;

in accordance with a determination that the current trie-node of the trie-structure has the reference to the first context trie-node corresponding to the value of the data bit, assigning the first context trie-node as a new current trie-node;

in accordance with a determination that the current trie-node of the trie-structure does not have the reference to the first context trie-node corresponding to the value of the data bit:

providing a new trie-node of the trie-structure;

assigning a reference of the current trie-node to the new trie-node;

traversing one or more trie-suffix-links associated with the current trie-node to identify a trie-node of the trie-structure having a reference to a second context trie-node corresponding to the value of the data bit;

providing a trie-suffix-link between the new trie-node and the second context trie-node;

copying a probability of the second context trie-node to the new trie-node; and assigning the second context trie-node as the new current trie-node;

predicting the next bit value in the bit stream using the probabilities of the current trie-node of the trie-structure; and

initiating a the data compression optimization procedure, wherein a set of weighting coefficients obtained by the compression optimization procedure is used as a part of a cryptographic key.