Digital Commons Network^™

Reinforcement Learning For Fuzzing Testing Techniques, Neil Dhillon, Tanmay Wadhwa

Defensive Publications Series

Generally, the present disclosure is directed to using machine learning to manage a trade-off between exploration and exploitation in a fuzzer system. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict inputs that will cause a system being tested to malfunction or crash based on input data provided to the system and output from the system.

Improving Appeal Of An Advertisement Based On Linguistic Trends, Donny Greenberg

Defensive Publications Series

Generally, the present disclosure is directed to improving appeal of an advertisement based on linguistic trends. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to generate fashionable advertisements based on a candidate advertisement and linguistic trends determined from sample material from one or more advertising platforms.

Configuring Alarm System Based On Time To Arrive At Appointment, Nick Felker

Defensive Publications Series

Generally, the present disclosure is directed to setting an alarm to alert a user based on an appointment. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a time to alert a user based on a user’s schedule and location.

Predictive Cryptocurrency Mining And Staking, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to determining the validity of a chain within a blockchain system. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict the likelihood that a block within a blockchain system will be verified based on characteristics of the block.

Predicting Computing Prices Dynamically Using Machine Learning, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to determining the cheapest datacenter for a computing task to be computed in. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict the cost of computing at a datacenter over a time interval based on time data, weather data, and/or grid statistics.

Predicting Delivery Time Of Components In A Supply Chain, Venki Raman, Sunil Reddy

Defensive Publications Series

Generally, the present disclosure is directed to predicting a delivery time of components in a supply chain. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a delivery time of a component based on enterprise data relating to the component.

Generating Travel Itineraries Based On User Interests, David Parkinson

Defensive Publications Series

Generally, the present disclosure is directed to generating a travel itinerary for a user based on the user’s interests. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict interest of a user and generate a travel itinerary based on user preferences and interests.

Identifying Hold State In An Automated Calling System, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to identifying that a caller is on hold or has been removed from hold when calling a calling system such as, for example, an automated calling system. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict that a hold state has changed (e.g. a caller has been placed on hold or has been taken off hold) based on call data from a call.

Machine Learning To Determine Type Of Vehicle, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to determining the vehicle type of a vehicle (e.g. for use in vehicle navigation). In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a vehicle type of a vehicle based on vehicle data relating to the operation and/or capacity of a vehicle.

Machine Learning To Identify Vehicle Maintenance Needs, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to identifying maintenance needs for a vehicle. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict vehicle maintenance needs for a vehicle based on vehicle data from the vehicle.

Identifying Energy Overconsumption Based On Ambient Noise Audio Signatures, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to identifying energy overconsumption in a device. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a device is in a state of energy overconsumption based on ambient noise data and electricity throughput.

Image Moderation Using Machine Learning, Dave Feltenberger, Rob Neuhaus

Defensive Publications Series

Generally, the present disclosure is directed to moderating images using machine learning. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a probability that an image will be accepted and/or rejected according to one or more rules of a submission service based on image data from the image.

Machine Learning For Accurate Battery Run Time Prediction, Liang Jia

Defensive Publications Series

Generally, the present disclosure is directed to using machine learning to manage state of charge of a battery. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict future energy consumption and optimal charging rate of a battery based on battery characteristics and expected user routine.

Removing Breathing Artifacts From Audio, N/A

Defensive Publications Series

Generally, the present disclosure is directed to removing breathing artifacts from audio. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to generate a clean audio sample based on audio data from the audio sample.

Configuring Music Variation In A Music Compilation, Thomas Price

Defensive Publications Series

Generally, the present disclosure is directed to determining an optimal level of music variation in an assortment of music such as, for example, a music streaming service. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an optimal level of music variation based on historical level of music variation preferences and/or listening data.

Discovering Employment Listings From Imagery, Charles Armstrong

Defensive Publications Series

Generally, the present disclosure is directed to discovering employment listings from imagery. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to identify employment listings based on image data.

Stylizing Map Based On Examples Of Representative Styling, Joshua Schpok

Defensive Publications Series

Generally, the present disclosure is directed to stylizing a map based on one or more examples of representative styling. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict map styling rules based on one or more examples of representative styling.

Improving Advertisement Delivery In Video Streaming, Mary Witkowski, Marshall Witkowski

Defensive Publications Series

Generally, the present disclosure is directed to improving advertisement delivery based on the content of a video. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a non-intrusive location for an advertisement based on the content of a video.

Predicting Search Queries Based On Personal Routines, Greg Hartrell

Defensive Publications Series

Generally, the present disclosure is directed to a computing system for predicting user intent in operation of a device based on personal routines. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict user intent in operation of a device based on past and/or current operation of the device and routine of the user.

Allocating System Resources In Display Of Advertisements, Bin Liu, Jiefu Zheng

Defensive Publications Series

Generally, the present disclosure is directed to predicting whether bid data of a present ad query (e.g., the largest cost per thousand impressions provided by a bid responsive to the ad query) will satisfy a threshold (e.g., a threshold cost per thousand impressions) based on bid data from one or more previous ad queries and/or characteristics of the present ad query. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict whether bid data of an ad query will satisfy a threshold based on bid …

Predicting Categorical Information For New Content In Application Recommender Systems, Jun Hong, Xiao Zhang

Defensive Publications Series

Generally, the present disclosure is directed to predicting categorical information for newly created content in an application recommender system. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict categorical information of new applications in an applications marketplace based on screenshots of such new applications in the applications marketplace.

Advertising Using Augmented Reality, Catherine Cardinal, Chin Ngai Sze

Defensive Publications Series

Heads-up displays have been gradually implemented in automobiles. These displays can provide information to drivers, such as speed, temperature, remaining fuel, navigation, and more, while minimizing distraction. Some vehicle heads-up displays add turn by turn directions based on GPS information onto a real-time street view. This disclosure describes techniques to show advertisements to passengers in vehicles using windows capable of implementing augmented reality (AR) - AR-capable windows.

Verbrauchsoptimierung, Daniel Hoppe

Defensive Publications Series

FUEL CONSUMPATION OPTIMIZATION BY APPROXIMATION

OF IMPULATIONS IN PARTICLE FILTER FOR

SPARK IGNITION ENGINE CONCEPTS

Throttle Grip: Potentiometer With Integrated, Daniel Hoppe

Defensive Publications Series

THROTTLE GRIP: POTENTIOMETER WITH INTEGRATED

RESET SPRING IN HANDLE VARIANT 2

Interface For Exploring Videos, Christopher Pentoney, Rodrigo De Oliveira, Anshuman Kumar, Mary Cassin, Joshua Lewandowski

Defensive Publications Series

Searching and browsing for videos is a common task. Search engines attempt to provide a quality user experience by showing the most relevant results at the top of a results page. However, efficiency and relevancy can sometimes be less of a concern, e.g., when browsing for interesting content, such as videos, e.g., without entering a search query. When users browse or search for videos, showing a hierarchical list of search results does not enable easy discovery of other content that they may enjoy.

This disclosure describes techniques to improve user enjoyment when browsing videos, e.g., directly, or after performing a …

Automated Fiber Network Monitoring, Bravishma Narayan

Defensive Publications Series

Telecommunication infrastructure providers, e.g., fiber network providers, benefit from accurate knowledge of the network footprint. This disclosure describes techniques to automatically monitor fiber networks. Datasets of fiber network information are hosted on suitable platforms, and can be easily customized. Network footprint, including information about the handholes and their adjacent handholes, provided by the network planner is captured as a dataset. Network splicing data points are identified and stored per customer along with their duct/microduct and ring information as a dataset. A breadth first search technique is used to analyze the datasets and provide information about usage of pairs and cores. …

Product Or Service Research Assistant, Google Llc

Defensive Publications Series

This document is directed to a product or service research assistant that helps you choose which product to buy from the many options available. Although this document will focus on a product research assistant, each idea also applies to a research assistant user interface (UI) for services professionals, travel destinations, restaurants or other structured entities and should be considered in the scope.

A Method For Detecting Scheduled-Service Vehicles By Crowdsensing Of On-Board Beacons, Lin Chen, Alexander Fabrikant, Rachael Morgan

Defensive Publications Series

The technology detects scheduled-service vehicles using a model generated based on onboard beacon location data. A trajectory of the onboard beacon may be determined by correlating the beacon detection data with the location data of the detecting devices. The trajectory may be analyzed to derive features of the beacon trajectory that correlate with a periodic motion typical of scheduled-service vehicles. Such features may be used to train a machine capable of learning patterns from data to generate a model that classifies whether a beacon trajectory is that of a scheduled-service vehicle. Various beacon trajectories may be classified by the generated …

Utilizing User-Level Signals For Early Spam Processing, Vladimir Rychev, Varun Sharma

Defensive Publications Series

A mechanism for early detection of a spam item by utilizing user-level signals associated with the user uploading the content item in advance of spending significant resources on processing the content item is disclosed. Upon detection of initiation of a content item uploaded by a user on a content item service, user-level signals associated with the user uploading the content item may be retrieved. Based on the user-level signals, it may be determined that the content item is a likely spam item. Subsequently, limited processing maybe performed by converting the content item may into a single file and content analysis …

Intelligent Assistant For Music Practice, Marcos Calvo, Philip Koonce

Defensive Publications Series

Generally, the present disclosure is directed to techniques to automatically provide feedback and suggestions to musicians. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to provide real-time feedback to musicians based on audio and/or video of the musician playing music. The techniques of this disclosure use various input features, e.g., the musician’s practice piece; references from a database of musical scores, data from different sensors, e.g., microphones, cameras, etc. to analyze the musician’s playing and provide real-time feedback or suggestions for corrections to be made, …