Neural network system and uses thereof
First Claim
Patent Images
1. A neural network system, comprising:
- i) an input layer configured to accept N input signals;
ii) one or more “
i”
hidden layers,iii) at least one output layer;
iv) at least one neuron “
Ylayer”
within each layer, where “
layer”
is the layer defined as “
input”
, “
hiddeni”
or “
output”
; and
,v) one or more memory structures configured to;
a) store a recursive memory of input signals past, andb) allow for at least one time series prediction of a response,wherein the neural network is configured to include one or more input variables oriented trend analysis post processing algorithms which are configured to analyze one or more predicted outputs of the neural network system for one or more expected trends in predicted analyte values based on previous and current input data presented to the neural network system;
wherein one or more support/post processing algorithms are included in order to modify the neural network system predictive output such that an increased predictive accuracy is achieved; and
wherein the neural network includes a support/post processing algorithm comprising an input variable, or event, oriented trend analysis algorithm.
1 Assignment
0 Petitions
Accused Products
Abstract
A multifunctional neural network system for prediction which includes memory components to store previous values of data within a network. The memory components provide the system with the ability to learn relationships/patterns existent in the data over time.
-
Citations
25 Claims
-
1. A neural network system, comprising:
-
i) an input layer configured to accept N input signals; ii) one or more “
i”
hidden layers,iii) at least one output layer; iv) at least one neuron “
Ylayer”
within each layer, where “
layer”
is the layer defined as “
input”
, “
hiddeni”
or “
output”
; and
,v) one or more memory structures configured to; a) store a recursive memory of input signals past, and b) allow for at least one time series prediction of a response, wherein the neural network is configured to include one or more input variables oriented trend analysis post processing algorithms which are configured to analyze one or more predicted outputs of the neural network system for one or more expected trends in predicted analyte values based on previous and current input data presented to the neural network system; wherein one or more support/post processing algorithms are included in order to modify the neural network system predictive output such that an increased predictive accuracy is achieved; and wherein the neural network includes a support/post processing algorithm comprising an input variable, or event, oriented trend analysis algorithm. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20)
-
-
21. A neural network system, comprising:
-
i) an input layer configured to accept N input signals; ii) one or more “
i”
hidden layers,iii) at least one output layer; iv) at least one neuron “
Ylayer”
within each layer, where “
layer”
is the layer defined as “
input”
, “
hiddeni”
or “
output”
; and
,v) one or more memory structures configured to; a) store a recursive memory of input signals past, and b) allow for at least one time series prediction of a response, wherein one or more memory structures are included in both the input and hidden layers, wherein the neural network is configured to include one or more input variables oriented trend analysis post processing algorithms which are configured to analyze one or more predicted outputs of the neural network system for one or more expected trends in predicted analyte values based on previous and current input data presented to the neural network system; wherein one or more support/post processing algorithms are included in order to modify the neural network system predictive output such that an increased predictive accuracy is achieved; and wherein the neural network includes a support/post processing algorithm comprising an input variable, or event, oriented trend analysis algorithm. - View Dependent Claims (22, 23)
-
-
24. A neural network system, comprising:
-
i) an input layer configured to accept N input signals; ii) one or more “
i”
hidden layers,iii) at least one output layer; iv) at least one neuron “
Ylayer”
within each layer, where “
layer”
is the layer defined as “
input”
, “
hiddeni”
or “
output”
; and
,v) one or more memory structures configured to; a) store a recursive memory of input signals past, and b) allow for at least one time series prediction of a response; wherein the neural network system is configured for both real-time prediction and retrospective prediction; wherein one or more support/post processing algorithms are included in order to modify the neural network system predictive output such that an increased predictive accuracy is achieved; wherein the neural network includes one or more support/post processing algorithms selected from;
an adaptive analyte threshold based rate of change (ROC) algorithm and input variable, or event, oriented trend analysis algorithm;wherein the adaptive analyte threshold based ROC post processing algorithm is configured to track the ROC of analyte data presented to the neural network system; wherein, based on a current analyte value and ROC of current and previous analyte values, if the n predicted analyte values do not correlate with the ROC, the predicted output of the neural network system are modified via the post processing algorithm to increase predictive accuracy; and wherein, n predicted values generated are adjusted to coordinate with real-time ROC to enhance predictive accuracy via Equation [2],
PREDICTmod=PREDICTCGM+WROC·
ROCprediect·
Δ
t
Equation[2],wherein, PREDICTCGM is a vector of predicted CGM values with length n, WROC is a vector of length n of weights for weighting ROC values based-on the current real time value (threshold), ROCpredict is a vector of ROC values of length n estimated based on best linear, or nonlinear model of real-time ROC, Δ
t is a time duration between the two samples or a sampling rate, andPREDICTmod is a vector of modified (post-processed) predictions to increase accuracy based on trends in real-time ROC.
-
-
25. A neural network system, comprising:
-
i) an input layer configured to accept N input signals; ii) one or more “
i”
hidden layers,iii) at least one output layer; iv) at least one neuron “
Ylayer”
within each layer, where “
layer”
is the layer defined as “
input”
, “
hiddeni”
or “
output”
; and
,v) one or more memory structures configured to; a) store a recursive memory of input signals past, and b) allow for at least one time series prediction of a response; wherein the neural network system is configured for both real-time prediction and retrospective prediction; wherein one or more support/post processing algorithms are included in order to modify the neural network system predictive output such that an increased predictive accuracy is achieved; wherein the neural network includes one or more support/post processing algorithms selected from;
an adaptive analyte threshold based rate of change (ROC) algorithm and input variable, or event, oriented trend analysis algorithm;wherein the adaptive analyte threshold based ROC post processing algorithm is configured to track the ROC of analyte data presented to the neural network system; wherein, based on a current analyte value and ROC of current and previous analyte values, if the n predicted analyte values do not correlate with the ROC, the predicted output of the neural network system are modified via the post processing algorithm to increase predictive accuracy; and wherein the post processing includes an adaptive analyte threshold based ROC approach;
-
Specification