Abstract: Systems, methods and media are provided for training a snippet extractor to create snippets based on information extracted from published descriptions. In one example, a computer-implemented method includes creating, based on a non-RNN (Recurrent Neural Network) extraction technique performed on the published descriptions, a plurality of base models, each base model including one or more sample description summaries; evaluating the base models using an evaluation technique; selecting an optimum base model; developing a classification model using RNN extraction, the classification model based on description summaries contained in the optimum base model; and using the classification model to train the snippet extractor by machine learning.

Abstract: Provided herein in some embodiments is an artificial intelligence (“AI”) engine configured to work with a graphical user interface (“GUI”). The AI engine can include an architect module, instructor module, and learner module AI-engine modules. The GUI can be configured with a text editor and a mental-model editor to enable an author to define a mental model to be learned by an AI model, the mental model including an input, one or more concept nodes, and an output. The architect module can be configured to propose a neural-network layout from an assembly code compiled from a source code in a pedagogical programming language, the learner module can be configured to build the AI model from the neural-network layout, and the instructor module can be configured to train the AI model on the one or more concept nodes.

Abstract: Provided herein in some embodiments is an artificial intelligence (“AI”) engine configured to operate with multiple user interfaces to accommodate different types of users solving different types of problems with AI. The AI engine can include AI-engine modules including an architect module, an instructor module, and a learner module. An assembly code can be generated from a source code written in a pedagogical programming language. The architect module can be configured to propose a neural-network layout from the assembly code; the learner module can be configured to build the AI model from the neural-network layout; and the instructor module can be configured to train the AI model built by the learner module. The multiple user interfaces can include an integrated development environment, a web-browser interface, or a command-line interface configured to enable an author to define a mental model for the AI model to learn.

Abstract: The present invention relates to unsupervised, supervised and reinforced learning via spiking computation. The neural network comprises a plurality of neural modules. Each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. An interconnection network comprising a plurality of edges interconnects the plurality of neural modules. Each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.

Abstract: Example embodiments for efficient neuromorphic population coding are described. In one case, individual instances of input stimuli are evaluated using a set of feature encoding units to generate a population of encoded feature values. The population of encoded values for each of the individual input stimuli are arranged into a population code matrix. The population code matrix is factorized into a basis element matrix and a contribution coefficient matrix based on a number of basis vectors, where the number of basis vectors is selected to balance sparseness in the basis element matrix and reconstruction error of the population code matrix from the basis element matrix and the contribution coefficient matrix. The embodiments are compatible with neuromorphic hardware and can achieve compact representation of high-dimensional data, infer latent variables in the data, and defer processing to an off-line training phase to save time during real-time data capture and evaluation.

Abstract: According to one embodiment, a method for generating a plurality of candidate visualizations. The method may include receiving a scenario description. The method may also include collecting a plurality of expert data using a training system based on the received scenario description. The method may further include generating at least one predictive model based on the collected plurality of expert data in order to execute the at least one generated predictive model during an application of a plurality of genetic algorithms.

Abstract: Predefined relation items on paths traversing predefined entities of a knowledge base are collected and feature sets are assembled from the collected relation items. A classifier is computed for the feature sets and a relation score of a query pair of the entities is computed using the classifier.

Abstract: A system for determining current and potential ranges of a vehicle includes a sensor that is used to detect energy data corresponding to an amount of energy remaining in the vehicle, such as fuel in a gasoline engine or fuel cell or a state of charge (SOC) of a battery. The system also includes a processor that can determine a current range of the vehicle in at least one direction based on a current status of the vehicle and the remaining energy levels. The processor can also determine a potential range of the vehicle that can be achieved if at least one user-controllable factor is changed from the current status of the vehicle. The system also includes an output device for outputting the current range of the vehicle and the potential range of the vehicle. The system may show fuel and/or recharging stations within the current and potential ranges.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving a message authored by a user, determining, using a first classifier, that the message contains at least a first word describing positive or negative sentiment and, based thereon, extracting, using a first feature extractor, one or more features of the message, wherein each feature comprises a respective word or phrase in the message and a respective weight signifying a degree of positive or negative sentiment, and determining, using a second classifier that uses the extracted features as input, a score describing a degree of positive or negative sentiment of the message, wherein the first feature extractor was trained with a set of training messages that each was labeled as having positive or negative sentiment.

Abstract: A device may receive issue resolution information, associated with a cognitive model, including an item of issue resolution information that describes an issue and a resolution to the issue. The device may assign the item of issue resolution information to a domain hierarchy, where the assigning is associated with a first user. The device may generate a question and an answer corresponding to the item of issue resolution information, where the generating of the question and the answer is associated with a second user. The device may approve the question and the answer, where the approving is associated with a third user. The device may generate a question/answer (QA) pair for the question and the answer. The device may create a data corpus including the QA pair, and provide the data corpus to cause the cognitive model to be trained based on the data corpus.

Abstract: A wearable apparatus may include a sensor configured to identify a characteristic of the wearable apparatus. The wearable apparatus may also include an activity tracker configured to identify an activity of a tracked individual. The wearable apparatus may further include a memory. The wearable apparatus may include a processor operatively coupled to the memory. The processor may be configured to receive portable autonomous decision logic via a network and execute the portable autonomous decision logic. When executing the autonomous decision logic, the processor is configured to receive the characteristic via the sensor, receive the activity via the activity tracker; determine a state of the tracked individual based on at least one of the characteristic or the activity; and generate an output based on the state of the tracked individual.

Abstract: A method for determining national crop yields during the growing season may be accomplished using a server computer system that receives, via a network, agricultural data records that are used to forecast a national crop yield for a particular year. Within the server computer system weather index calculation instructions receive one or more agricultural data records that represent observed agricultural data points for a specific geo-location at a specific time. The observed agricultural data points include, but are not limited to, observed temperature records, soil moisture records, and precipitation records for the specific geo-location. The weather index calculation instructions then calculate one or more weather index values from the one or more agricultural data records. Weather index values represent crop stress that may affect the crop yield of particular plants.

Abstract: A method of incident root cause analysis in an information technology (IT) system, wherein upon occurrence of an incident collecting changes to configuration items and/or system parameters on computer stations during a predetermined time prior to the incident, calculating a change lifetime for each of the collected changes, comparing the change lifetime to the time of occurrence of the incident to determine if the lifetime of the change is still valid, marking a probability value of zero for occurrence of the incident as a result of the change for changes with an expired lifetime value at the time of the incident, otherwise estimating a risk profile and calculating from it a probability value for occurrence of the incident as a result of the change, sorting the changes according to the probability value, and selecting a predetermined number of changes having the highest probability values for root cause analysis.

Abstract: A technique relates to an assembly for a quantum computing device. A quantum bus plane includes a first set of recesses. A readout plane includes a second set of recesses. A block is positioned to hold the readout plane opposite the quantum bus plane, such that the first set of recesses opposes the second set of recesses. A plurality of qubit chips are included where each has a first end positioned in the first set of recesses and has a second end positioned in the second set of recesses.

Abstract: In some aspects, a mesoscopic system is used to generate entanglement, for example, on a pair of qubits. In some implementations, the mesoscopic system includes a first spin, a second spin and multiple other spins. The initial state of the first and second spins can be separable (non-entangled) states. Measurement outcomes are obtained by performing projective measurements on the mesoscopic system. Based on the measurement outcomes, an entangled state of the first and second spins is detected. The entangled state is transferred from the first and second spins to the first and second qubits.

Abstract: A method predicts the fall risk of a user based on a machine learning model. The model is trained using data about the user, which may be from wearable sensors and depth sensors, manually input by the user, and received from other types of sources. Data about a population of users and data from structured tests completed by the user can also be used to train the model. The model uses features and motifs discovered based on the data that correlate to fall risk events to update fall risk scores and predictions. The user is provided a recommendation describing how the user can reduce a predicted fall risk for the user.

Abstract: Described herein is a system and method for addition of content to a question answer (QA) system. The method may comprise retrieving, from a template store of the QA system, a template comprising one or more data fields and one or more metadata fields. The method may further comprise populating a passage field with passage information, wherein the passage field is one of the one or more data fields. The method may further comprise populating a training field with training information, wherein the training field is one of the one or more metadata fields. The method may further comprise ingesting, into a corpus of the QA system, a completed template comprising a populated passage field and a populated training field into the corpus.

Abstract: One embodiment provides a method for generating description text for a desired application using a machine classifier trained using other applications, the method including: utilizing at least one processor to execute computer code that performs the steps of: obtaining description text and at least one application component for an application; identifying at least one application characteristic from the at least one application component; associating at least one word expression contained within the description text to at least one identified application characteristic; training the machine classifier, wherein the training comprises identifying similar application characteristics within the at least one application component and identifying, based upon the associating, a condition of the at least one application characteristic including the at least one word expression; obtaining at least one application component for the desired application; identifying, using the trained machine classifier, at least one app

Abstract: Systems, methods, and computer media for machine learning through a symbolic, parallelized stochastic gradient descent (SGD) analysis are provided. An initial data portion analyzer can be configured to perform, using a first processor, SGD analysis on an initial portion of a training dataset. Values for output model weights for the initial portion are initialized to concrete values. Local model builders can be configured to perform, using an additional processor for each local model builder, symbolic SGD analysis on an additional portion of the training dataset. The symbolic SGD analysis uses a symbolic representation as an initial state for output model weights for the corresponding portions of the training dataset. The symbolic representation allows the SGD analysis and symbolic SGD analysis to be performed in parallel. A global model builder can be configured to combine outputs of the local model builders and the initial data portion analyzer into a global model.

Abstract: A machine learning model is trained by defining a scenario including models of vehicles and a typical driving environment. A model of a subject vehicle is added to the scenario and sensor locations are defined on the subject vehicle. A perception of the scenario by sensors at the sensor locations is simulated. The scenario further includes a model of a lane-splitting vehicle. The location of the lane-splitting vehicle and the simulated outputs of the sensors perceiving the scenario are input to a machine learning algorithm that trains a model to detect the location of a lane-splitting vehicle based on the sensor outputs. A vehicle controller then incorporates the machine learning model and estimates the presence and/or location of a lane-splitting vehicle based on actual sensor outputs input to the machine learning model.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning using a partitioned reinforcement learning input state space (RL input state space). One of the methods includes maintaining data defining a plurality of partitions of a space of reinforcement learning (RL) input states, each partition corresponding to a respective supervised learning model; obtaining a current state representation that represents a current state of the environment; for the current state representation and for each action in the set of actions, identifying a respective partition and processing the action and the current state representation using the supervised learning model that corresponds to the respective partition to generate a respective current value function estimate; and selecting an action to be performed by the computer-implemented agent in response to the current state representation using the respective current value function estimates.

Abstract: A method of inverse reinforcement learning for estimating cost and value functions of behaviors of a subject includes acquiring data representing changes in state variables that define the behaviors of the subject; applying a modified Bellman equation given by Eq. (1) to the acquired data: q(x)+gV(y)?V(x)=?1n{pi(y|x))/(p(y|x)} (1) where q(x) and V(x) denote a cost function and a value function, respectively, at state x, g represents a discount factor, and p(y|x) and pi(y|x) denote state transition probabilities before and after learning, respectively; estimating a density ratio pi(y|x)/p(y|x) in Eq. (1); estimating q(x) and V(x) in Eq. (1) using the least square method in accordance with the estimated density ratio pi(y|x)/p(y|x), and outputting the estimated q(x) and V(x).

Abstract: A first indication from a user is received. The indication includes a task to be performed using at least one application programming interface. A machine learning model is determine. At least one application programming interface is determined using the machine learning model and the request. The at least one application programming interface is provided to the user.

Abstract: Systems and methods for executing an unsupervised feature selection algorithm on a processor which directly embeds feature selection into a clustering algorithm using sparse learning are disclosed. The direct embedding of the feature selection, via sparse learning, reduces storage requirement of collected data. In one method, unsupervised feature selection may be accomplished through a removal of redundant, irrelevant, and/or noisy features of incoming high-dimensional data.

Abstract: Provided herein in some embodiments is an artificial intelligence (“AI”) engine hosted on one or more servers configured to cooperate with one or more databases including one or more AI-engine modules. The one or more AI-engine modules include an architect module configured to propose an AI model from an assembly code. The assembly code can be generated from a source code written in a pedagogical programming language describing a mental model of one or more concept modules to be learned by the AI model and curricula of one or more lessons for training the AI model on the one or more concept modules in one or more training cycles. The AI engine can be configured to instantiate a trained AI model based on the one or more concept modules learned by the AI model in the one or more training cycles.

Abstract: An AI database hosted on cloud platform is configured to cooperate with a search engine and an AI engine. The AI database stores and indexes trained AI objects and its class of AI objects have searchable criteria. The AI database cooperates with the search engine to utilize search criteria supplied from a user, from either or both 1) via scripted software code and 2) via data put into defined fields of a user interface. The search engine utilizes the search criteria in order for the search engine to retrieve one or more AI data objects that have already been trained as query results. The AI database is coupled to an AI engine to allow any of reuse, reconfigure ability, and recomposition of the one or more trained AI data objects from the AI database into a new trained AI model.

Abstract: The multiple independent processes run in an AI engine on its cloud-based platform. The multiple independent processes are configured as an independent process wrapped in its own container so that multiple instances of the same processes can run simultaneously to scale to handle one or more users to perform actions. The actions to solve AI problems can include 1) running multiple training sessions on two or more AI models at the same time, 2) creating two or more AI models at the same time, 3) running a training session on one or more AI models while creating one or more AI models at the same time, 4) deploying and using two or more trained AI models to do predictions on data from one or more data sources, 5) etc. A service handles scaling by dynamically calling in additional computing devices to load on and run additional instances of one or more of the independent processes as needed.

Abstract: Various embodiments described herein provide for methods and systems of providing related information from a data source. For example, a method may include responding to a natural language query with a natural language response that includes and describes data related to the natural language query. The related data may correspond to the numerical data or text discovered from various data sources. Further, a database trained with a machine-learning algorithm may be utilized to identify time series related data that is associated with the natural language query and that is used within the generated natural language response. The methods and systems described herein may be utilized by a message bot when responding to questions posed by an online user during a chat session.

Abstract: A feature calculation unit 81 calculates a feature that is likely to influence cancellation by a user based on a communication state log that indicates a communication state of a base station when the user has been engaged in communication or making a call. A learning device 82 learns a model representing a behavioral characteristic of the user by using the calculated feature as an explanatory variable. A prediction device 83 predicts the behavioral characteristic of the user using the feature generated from the communication state log and the model.

Abstract: This invention of a revenue management willingness to pay higher fares indicator, provides a method to define in a computer based revenue management system, levels of willingness to pay higher fares, used to improve the forecasting of potential revenue and improve the availability controls defined by the revenue management system. Where current revenue management systems use booked fare class information as basis for forecasting revenue, when demand with a higher willingness to pay higher fares books in a low fare class, this information is not recognized as only the (lower) booked fare class information is stored and used. The method proposed defines levels of willingness to pay higher fares based on booking record characteristics without directly using the actual booked fare class and storing this information for usage in further steps in forecasting and defining availability controls by the revenue management system.

Abstract: A computer-implemented method for managing services is disclosed, which comprises providing to a service provider a notification of a service reservation by one or more customer entities, continuously estimating a time of arrival based on a current geolocation of the one or more customer entities, providing to the service provider an indication of the estimated time of arrival, notifying the service provider of an arrival of the one or more customer entities at the location of the service provider based on the current geolocation and the location of the service provider, and responsive to the arrival at the location of the service provider, automatically coupling the one or more customer entities to a communication network of the service provider to confirm the service reservation.

Abstract: An approach is provided for facilitating reservations of resources for a shorter time period (e.g., intra-day booking) than a traditional daily reservation and/or for extending reservations beyond the limits of traditional reservation times (e.g., check-in, check-out, etc.) The approach includes specifying an intra-day booking period for a resource, wherein the intra-day booking period is a period less than a daily booking period for the resource. The approach also includes interfacing with a channel manager platform to provide for an intra-day reservation of the resource based on the intra-day booking period, wherein the channel manager platform is configured to manage the resource. The approach further includes initiating the intra-day reservation of the resource based on a user intra-day reservation request.

Abstract: A method and system to provide an on-demand real-time virtual travel service to a traveler utilizing a virtual travel management system. The virtual travel management system provides virtual travel demonstration of geographical locations to the traveler without having to be physically present at those destinations. A proxy is equipped with a proxy communication device and a portable audio-video monitoring and transmitting device. The traveler is equipped with a traveler communication device and a multimedia receiving device. The proxy communication device and the traveler communication device are installed with a virtual travel management software application. The audio-video monitoring and transmitting device transmits a real-time demonstration of picture and sound from a proxy location to the multimedia receiving device in response to the traveler's virtual travel service real-time commands.

Abstract: A system and method for remote and anonymous check-in at a lodging facility enables a guest to check in to the lodging facility by remotely making a reservation to stay at the lodging facility for a predetermined duration, and anonymously entering and leaving the room. The duration of stay may include only a few hours to increase turnover for the lodging facility and flexibility to the guest. The guest selects a themed room. The guest remotely pays for the themed room through a software application on a smart phone. The guest anonymously checks in to the lodging facility. A kiosk provides an access mechanism for the selected themed room to the guest. The guest anonymously enters the themed room through a discrete route, such as a covered garage. The guest anonymously returns the access mechanism for the themed room, and remotely checks out from the lodging facility.

Abstract: Methods, autonomous vehicles, and servers for passenger service in an autonomous vehicle are provided. A method includes receiving at least one signal indicating a status of at least one passenger of the autonomous vehicle. The method further includes processing the at least one signal to determine whether the at least one passenger has taken action consistent with completion of the passenger service reservation in the absence of an action that is inconsistent with completion of the passenger service reservation. The method yet further includes completing the reservation based on the processing the at least one signal. A server includes a processor and a non-transitory computer readable medium storing instructions that configure the server for performing the method.

Abstract: A methods and a server for operating a vehicle are provided. A method includes selecting an identifier that is associated with a vehicle reservation for passenger service in the vehicle. The method further includes initiating a pick-up portion of the vehicle reservation for making the vehicle available to a passenger. The method further yet includes displaying the identifier at the vehicle during the pick-up portion of the passenger service. The server includes a processor and a non-transitory computer readable medium storing instructions that configure the server for performing the method.

Abstract: Disclosed is a system that may distribute the profit obtained through translation and publication by allowing a platform operator, a writer, a translator, and a portal site to cooperate with each other. The inventive concept relates to a system that allows an investment of an investor in a work determined to have a high possibility of being sold. A contract letter of commitment for distribution of a profit made by a management company, a writer, a translator, and a portal site is written, stored, and managed.

Abstract: A computer-implemented method is provided for providing at least one analytics package to a process mining system, wherein the processor is provided with an event log comprising process data of business processes, the process data comprising at least one process element and the process element comprising at least one process step. The event log is stored according to a predetermined data structure comprising at least a first attribute for storing a unique identifier of the process element, a second attribute for storing an identifier of the process step, and a third attribute for storing an order of the process steps. The processor is further provided with auxiliary data and a data model, and the method comprises creating, based on the data model, at least one analytics package.

Abstract: A method for optimizing risks in supply chain networks is disclosed. The method includes categorizing, via a risk optimizing device, contextually relevant keywords derived from a user query into a risk category selected from a plurality of risk categories; identifying, via the risk optimizing device, a risk in the supply chain network based on the contextually relevant keywords and the risk category; creating, via the risk optimizing device, a plurality of risk association rules representative of interdependencies of the risk with at least one associated risk; assigning, via the risk optimizing device, priority to each of the plurality of risk association rules based on impact of interdependent risks within corresponding risk association rules; and optimizing a risk association rule assigned high priority within the plurality of risk association rules by removing the risk or one of the at least one associated risk from the risk association rule.

Abstract: A computer implemented method and system for managing an organization in a virtual social network environment provides an organization management platform accessible to employees and clients over a communication network. The organization management platform acquires target goal criteria of the organization via a graphical user interface over the communication network, and creates multiple organizational groups based on the target goal criteria. The organization management platform registers the employees with one or more organizational groups selected by the employees. The organization management platform acquires one or more tasks and rules for performing the tasks that enable the employees to fulfill the target goal criteria.

Abstract: A software system is disclosed. A one or more computers comprises a one or more input devices, an address space, a processor, a memory and a communication hardware. Said one or more computers comprises said software system. Said one or more computers comprises a first computer and a server. Said software system comprises a device application and a server application. Said server comprises said server application and a system database. Said first computer comprises said device application. Said system database comprises a plurality of user account records, a plurality of animal account records, a one or more event records, a one or more judging records, a plurality of animal competition records, a one or more competition records, a one or more registrant records and an event summary records. Said plurality of user account records is configured to track a plurality of users of said software system.

Abstract: According to examples, intelligent scheduling and work item allocation may include ascertaining work items, and classifying the work items by using classification rules to map each of the work items to a corresponding type of work item based on attributes associated with the work items to generate classified work items. Intelligent scheduling and work item allocation may include prioritizing the classified work items by using prioritization rules to determine a sequence of the classified work items based on the attributes and classification of the work items to generate prioritized work items. Intelligent scheduling and work item allocation may include scheduling the classified and prioritized work items by using scheduling rules to determine times of processing of the classified and prioritized work items, and allocating the classified and prioritized work items by using allocation rules to determine resources that are to process the classified and prioritized work items.

Abstract: A computer implemented method for mining supply chain information to produce supply chain graphs includes receiving by a computer a set of data; identifying a supplier, a commodity, and a customer from the set of data; generating a query comprising the identified data; determining the absence of any of the supplier, the commodity, or the customer from the set of data; if any items are determined to be absent from the set of data, substituting a placeholder for the missing item from the data set; sending the query; receiving a set of supply chain graph information; generating a supply chain graph signal based upon the set of supply chain graph information; and transmitting the supply chain graph signal. A computing device or system includes a processor an electronic memory; and a program for mining supply chain information to produce supply chain graphs stored in the electronic memory.

Abstract: A monitoring system configured to automatically detect that an amount of items in an environment has fallen below a threshold amount and in response automatically trigger a shipment of one or more additional items. The disclosed monitoring system addresses deficiencies of traditional ‘subscription orders’ by ensuring stock is only replenished when it falls below a user-specified threshold, rather than at periodic intervals whether required or not.

Abstract: A system, method and computer program product for enhancing on-premise order management systems (OMS) designed for fulfillment transactions with analytic and optimization technologies and services hosted in a shared multi-tenant software-as-a-service (SaaS) environment, such as a hybrid cloud. The computer-implemented method improves an order management system by leveraging a “punch-out” approach based on user exits to integrate with and augment currently implemented order management processing and transaction flows. Using the hybrid cloud, an entity may retain data such as its accumulated business, sales, test and other data, and then run analytical queries, which can scale to support distributed computing tasks.

Abstract: The system and methods described herein monitor transactions occurring in any large complex enterprise software systems for simple or compound events that meet certain defined rules. Once events occur that successfully meet the criteria set out in each rule, then proactive notifications are sent to designated users. This enables early detection of suspicious transactions and also quick response in real-time.

Abstract: In one implementation, a workflow system can include a storage engine and a merger engine. The storage engine maintains a restricted workflow part on a first storage resource and maintains a customizable workflow part on a second storage resource. The merger engine retrieves the restricted workflow part based on the product version and merge the restricted workflow part with the customizable workflow part associated with the restricted workflow part.

Abstract: In one embodiment, a method of benchmarking energy assets is disclosed. The method includes filtering asset data received from a plurality of energy assets based on constraints to generate filtered asset data; creating a plurality of data profiles using the filtered asset data based on profiling variables; identifying at least one benchmarking variable and at least one normalizing variable for the plurality of energy assets for a data profile in the plurality of data profiles; iteratively determining correlation between the at least one benchmarking variable and the at least one normalizing variable; and normalizing the at least one benchmarking variable using the at least one normalizing variables in response to determining the correlation to generate benchmarks for each of the at least one benchmarking variable.

Abstract: A method for collecting data and deriving performance metrics for one or more users within an observation platform environment is described. The method includes recognizing a user identity based on signals from the communications device; Capturing a user's voice message and the secondary signal indicative of the location of the communications device or the relative proximity of the communications device to other devices; Deriving statistical data comprising: primary statistics, secondary statistics, and higher-order statistics from the secondary signal and voice message; and Storing the user identity, the voice message, the location of the communications device, and the statistical data in a database forming a user record, wherein the user record is processed by an inference engine to derive performance information about the user.

Abstract: This disclosure provides for a social networking server that helps a member of a social networking service find a bridge between his or her current skill set and skills associated with a job listed on the social networking service. The social networking server forms a skills graph for the members of the social networking service, where nodes of the skills graph correspond to skills possessed by the members. The nodes are collected into pairs, where an edge between a selected pair is established based on a confidence measure that a given member possesses each skill of the selected pair. By traversing a path through the nodes of the skills graph, the social networking service helps a member identify intermediary skills that the member could learn to help bridge a gap between his or her current skill set and skills associated with a job opening.