Information processsing apparatus, information processing method, and program
An information processing apparatus for configuring algorithms is disclosed. The information processing apparatus includes an algorithm configuring section that configures an algorithm for performing discrimination on an input signal by using a genetic search technique. The algorithm includes feature extraction expressions and an information estimation expression represented by a combination of the feature extraction expressions. The information processing apparatus also includes a tradeoff analyzing section that determines pareto optimal solutions by optimizing the algorithm with respect to evaluation indices by performing tradeoff analysis on the basis of the algorithm. In addition, the information processing apparatus includes a storage for storing the algorithm.
1. An information processing apparatus comprising:
an algorithm configuring section that configures an amount-of-feature extraction algorithm that determines whether an input signal has a particular characteristic by using a genetic search technique, the algorithm comprising:
feature extraction expressions that specify:
a type of the input signal, the type indicating a type of data representing the input signal; and
operations to be performed on the input signal; and
an information estimation expression including a linear combination of the feature extraction expressions, wherein the information estimation expression uses first-order values output from the feature extraction expressions to estimate information indicating features of the input signal;
a tradeoff analyzing section that generates pareto optimal solutions by selecting information estimation expressions having maximum values of evaluation indices;
an optimum algorithm determining section that selects, from the pareto optimal solutions, an optimum algorithm that matches a requested condition of the evaluation indices; and
a storage for storing the algorithm.
2. The information processing apparatus according to claim 1 , wherein performing the operations specified by each feature extraction expression on the input signal produces a scalar value.
3. The information processing apparatus according to claim 1 , wherein the tradeoff analyzing section generates pareto optimal solutions in an initial state by using algorithms having different numbers of feature extraction expressions, the algorithms being created by deleting the feature extraction expressions one by one from the algorithm configured by the algorithm configuring section.
4. The information processing apparatus according to claim 3 , wherein the tradeoff analyzing section updates the pareto optimal solutions by randomly changing use or not-use of each feature extraction expression in the pareto optimal solutions in the initial state.
5. The information processing apparatus according to claim 1 , wherein the tradeoff analyzing section uses, as a pareto optimal solution in an initial state, the algorithm configured by the algorithm configuring section.
6. The information processing apparatus according to claim 1 , wherein the optimum-algorithm determining section determines, of the pareto optimal solutions, the optimum algorithm on a basis of requested processing time and accuracy.
7. The information processing apparatus according to claim 1 , further comprising an evaluation-value calculating section that determines evaluation values of the feature extraction expressions in the information estimation expression, wherein the algorithm configuring section updates the feature extraction expressions in the information estimation expression, on a basis of the determined evaluation values of the feature extraction expressions.
8. The information processing apparatus according to claim 7 wherein when the algorithm configuring section configures multiple information estimation expressions as algorithms, a total value of contribution rates of the same feature extraction expressions in the information estimation expressions is used as the evaluation value of the corresponding feature extraction expression.
9. A computer-implemented information processing method comprising:
configuring, using a processor, an amount-of-feature extraction algorithm that determines whether an input signal has a particular characteristic by using a genetic search technique, the algorithm comprising:
feature extraction expressions that specify:
a type of the input signal, the type indicating a type of data representing the input signal; and
operations to be performed on the input signal; and
an information estimation expression including a linear combination of the feature extraction expressions, wherein the information estimation expression uses first-order values output from the feature extraction expressions to estimate information indicating features of the input signal; and
generating pareto optimal solutions by selecting information estimation expressions having maximum values of evaluation indices; and
selecting, from the pareto optimal solutions, an optimum algorithm that matches a requested condition of the evaluation indices.
10. The computer-implemented method according to claim 9 , wherein optimizing includes generating pareto optimal solutions in an initial state by using algorithms having different numbers of feature extraction expressions, the algorithms being created by deleting the feature extraction expressions one by one from the algorithm.
11. The computer-implemented method according to claim 9 , wherein in an initial state, the algorithm configured in the configuring step is used as-as a pareto optimal solution.
12. The computer-implemented method according to claim 9 , wherein determining pareto optimal solutions includes updating the pareto optimal solutions by randomly changing use or not-use of each feature extraction expression in the pareto optimal solutions in the initial state.
13. The computer-implemented method according to claim 9 , wherein the optimum algorithm is determined on a basis of requested processing time and accuracy.
14. The computer-implemented method according to claim 9 , further comprising:
determining evaluation values of the feature extraction expressions in the information estimation expression, and
updating the feature extraction expressions in the information estimation expression, on a basis of the determined evaluation values of the feature extraction expressions.
15. The computer-implemented method according to claim 14 , wherein in the configuring step, a total value of contribution rates of the same feature extraction expressions in the information estimation expressions is used as the evaluation value of the corresponding feature extraction expression.
16. A non-transitory computer-readable medium storing instructions for causing a computer to execute operations comprising:
configuring an amount-of-feature extraction algorithm that determines whether an input signal has a particular characteristic by using a genetic search technique, the algorithm comprising:
feature extraction expressions that specify:
a type of the input signal, the type indicating a type of data representing the input signal; and
operations to be performed on the input signal; and
an information estimation expression including a linear combination of the feature extraction expressions, wherein the information estimation expression uses first-order values output from the feature extraction expressions to estimate information indicating features of the input signal;
generating pareto optimal solutions by selecting information estimation expressions having maximum values of evaluation indices; and
selecting, from the pareto optimal solutions, an optimum algorithm that matches a requested condition of the evaluation indices.