International Business Machines Corporation

Sparkbeyond Ltd.

Amir Ronen

Ron Karidi

Meir Maor

Sagie Davidovich

Ehud Aharoni

Eran Raichstein

Tamer Salman

Dan S Pelleg

Yehuda Naveh

Shmuel Regev

Ron Peleg

Seema Nagar

Amit Anil Nanavati

Ramasuri Narayanam

Steven A Keller

Dorit Baras

Denis Wainshtein

Jason Hardy

Avishay Livne

Yuval Peleg

A classifier is computed as follows. For a first set of values of primary field(s) of primary data instances of a labeled primary training dataset, a second set(s) of secondary fields of unclassified second data instances of secondary dataset(s) is identified. First set of values are matched to corresponding values in respective secondary field(s), and linked to other secondary fields of respective secondary data instance(s) of the respective matched secondary field. A set of classification features is generated. Each including: (i) condition(s), and (ii) a value selected from the linked other secondary fields of the respective secondary data instance(s) of the respective matched secondary field(s). The respective classification feature outputs a binary value computed by the condition(s) that compares between the value selected from the other linked secondary fields and a new received data instance. A classifier is computed according to a selected subset of the classification features.

Systems and methods for secondary knowledge utilization in machine learning

Methods, products and apparatus are provided for hypotheses generation using searchable unstructured data corpus. In one method, a query is generated based on at least one attribute of at least one instance in a dataset. The query is provided to a search engine searching in an unstructured data corpus. An hypothesis for the database is based on a new attribute whose value is defined based on the one or more results. Another method comprises obtaining a set of keywords from a plurality of hypotheses extracted from a database. A query is generated based on an attribute of an instance in the dataset, where the attribute corresponds to an hypothesis. A search engine executes the query to provide results which are used to augment an instance with a new attribute, where a value of the new attribute is computed based on the one or more results.

Hypotheses generation using searchable unstructured data corpus

There is provided a method for searching an unstructured dataset with a query, comprising: receiving a query comprising a value for a first token of a triplet, and a value for a relation term defining a relationship between the first token and a second token of the triplet, wherein the second token is defined as a variable element set with an undefined value, creating a plurality of enhanced queries for the query, each one of the plurality of enhanced queries including variations of the relation term, providing the plurality of enhanced queries for search by a search engine on at least one dataset of unstructured text-based data, receiving a plurality of documents in response to the search, analyzing the plurality of documents for extracting at least one value for the variable element of the triplet, and providing the at least one value for the variable element.

Systems and methods for searching an unstructured dataset with a query

In one embodiment, payoffs to an actor during the strategic game according to a solution concept is simulated a predetermined number of times. A machine learning tool is used to predict a payoff based on the predetermined number of payoffs according to the solution concept, wherein the machine learning tool is trained with plurality of pairs, each pair comprising a set of simulated payoffs and an actual payoff, wherein the simulated payoffs are predetermined number of payoffs determined in accordance with the solution concept. In another embodiment, a process tree representative of a strategic game involving actors is obtained. The process tree may comprise computational nodes, and leaf nodes which define payoff based on computations of the computational nodes. The process tree is reduced to a game tree representing the strategic game by simulating routes in the process tree.

System and method for decision making in strategic environments

There is provided a method for generating features for use in an automated machine learning process, comprising: receiving a first training dataset comprising unclassified raw data instances each including a set of objects of arbitrary types; applying a function to each data instance to calculate a set of first results; generating a set of classification features each including the function for application to a newly received data instance to calculate a second result, and a condition defined by a respective member of the set of first results applied to the second result; applying each classification feature to each instance of an unclassified second training dataset to generate a set of extracted features; selecting a subset of pivotal classification features from the set of classification features according to a correlation requirement between classification variable(s) and each respective member of the set of extracted features; and documenting the subset of pivotal features.

System and method for automatic generation of features from datasets for use in an automated machine learning process

Detecting computer anomalies by determining probabilities of encountering call stack configurations at various depths, the call stacks being associated with software application instances on computers having the same operating system, where snapshots of the call stacks are recorded on the computers responsive to detecting predefined software application events, determining entropies of call stack configurations at various call stack depths using their associated probabilities, determining stack frame rarity scores of call stack configurations at various depths based on their associated stack frame entropies in accordance with a predefined rarity function, determining a call stack rarity score of any given call stack configuration as the maximum stack frame rarity score of the given configuration, and detecting an anomaly associated with any given one of the computers where any of the snapshots recorded on the given computer is of a call stack whose call stack rarity score meets a predefined anomaly condition.

Using call stack snapshots to detect anomalous computer behavior

There is provided a computer-implemented method of identifying anomalous entities in a dataset, comprising: selecting a subset of training entities from entities of at least one dataset; determining dummy tuplets of entities in the subset by applying a permutation function on real tuplets, wherein the real tuplets represent original and normal data of the at least one dataset, wherein the dummy tuplets represent anomalous data based on artificially created data not found in the original and normal at least one dataset, each one of the real tuplets and dummy tuplets comprises at least two of the training entities; analyzing the dummy tuplets and the real tuplets to identify at least one predefined characteristic relation that statistically differentiates between the real tuplets and the dummy tuplets according to a distinguishing requirement; and outputting the identified at least one predefined characteristic relation to identify a normal entity and/or an anomalous entity.

Systems and methods for detection of anomalous entities

A method comprising calculating for each agent, an average quality of tasks that were completed in the past by the agent; allocating tasks to the agents, wherein said allocating comprises selecting an agent to perform a task, the selection is based on the average quality of the agent; in response to the agent completing the task, computing a reward for the agent, wherein the reward is calculated according to a total contribution of the agent to the system by completing the task; whereby biasing said allocating to prefer allocating tasks to a first agent over a second agent, if a quality of the first agent is greater than a quality of the second agent, wherein said biasing is not dependent on prior knowledge of the qualities. Optionally, the agents choose whether or not to perform a task and an agent's quality affects the contributions of the agent performing tasks.

System, method and product for task allocation

Techniques, a system and an article of manufacture for determining potential enterprise partnerships. A method includes constructing an interaction graph for a set of multiple enterprises based on information pertaining to each of the enterprises and information pertaining to transactions conducted among the enterprises, and determining at least two candidate partnership teams among the set of enterprises based on at least one attribute of the graph, wherein each candidate partnership team comprises two or more of the enterprises. The method also includes computing a compatibility score for each pair of enterprises in each candidate partnership team based on at least one enterprise parameter, computing a compatibility score for each candidate partnership team based on an aggregation of the compatibility score for each pair of enterprises in each candidate partnership team, and outputting a suggested candidate partnership team based on selection of the highest compatibility score among the candidate partnership teams.

Determining potential enterprise partnerships

A method, apparatus, and computer program product for generating a social network data structure, the method comprising: receiving a corpus comprising one or more communication indications for one or more customers, each communication indication indicating start time and end time of a communication of the customer; and generating a social network data structure indicating connections between customers based upon the start time and end time indications of communication indication comprised in the communication indications.

Extracting social relations from calling time data

Read-ahead of data blocks in a storage system is performed based on a policy. The policy is stochastically selected from a plurality of policies in respect to probabilities. The probabilities are calculated based on past performances, also referred to as rewards. Policies which induce better performance may be given precedence over other policies. However, the other policies may be also utilized to reevaluate them. A balance between exploration of different policies and exploitation of previously discovered good policies may be achieved.

Growing community of inventors

Haifa, Israel

Amir Ronen