Sap Se

Other

Sap Ag

Ran Bittmann

Michael Kemelmakher

Leonid Bobovich

Lev Sigal

Roy Gelbard

Ilia Rutenburg

Hans-Martin Ramsl

Roman Talyansky

Itzhak Shoshan

Yuri Arshavski

Laurent Boué

Isaac Eliassi

Anna Fishbein

Avi Rokach

Tatiana Surazhsky

Matan Mizrahi

Yan Vulich

Iuliia Drozdova

Ahmad Haj Ali

Anna Gurayevskaya

Systems, methods, and computer-readable media are disclosed for visual labeling of training data items for training a machine learning model. Training data items may be generated for training the machine learning model. Visual labels, such as QR codes, may be created for the training data items. The creation of the training data item and the visual label may be automated. The visual labels and the training data items may be combined to obtain a labeled training data item. The labeled training data item may comprise a separator to distinguish the training data item from the visual label. The labeled training data item may be used for training and validation of the machine learning model. The machine learning model may analyze the training data item, attempt to identify the training data item, and compare the identification against the embedded label.

Visual labeling for machine learning training

Some embodiments provide a program that queries a database for a subset of a plurality of records in the database. Each record in the plurality of records includes a value for a field. The program further samples the subset of the plurality of records to identify a set of records in the subset of the plurality of records. The program also sorts the set of records based on the value for the field in each record in the set of records. The program further determines a first value for the field of a first record in the sorted set of records and a second value for the field of a second record in the sorted set of records forms a slope that is greater than or equal to a defined slope. The program determines a threshold value for the subset of the plurality of records based on the first record.

Determining threshold values based on sampled data

Some embodiments provide a non-transitory machine-readable medium that stores a program. The program queries a database for a subset of a plurality of records in the database. Each record in the plurality of records includes a value for a first field and a second value for a second field. The program further normalizes the first value of the first field of each record in the subset of the plurality of records. The program also divides the subset of the plurality of records into a plurality of groups of records based on the second values of the second field. The program further generates a function for predicting a type of a particular record based on the value of the field of the particular record.

Predicting types of records based on amount values of records

Techniques for implementing blind spots into neural network models are disclosed. In some example embodiments, a computer-implemented method comprises obtaining an image captured within a field of view of an image capture device and comprising an object of a particular type occupying a particular position within the field of view, and determining a confidence value for the object based on the particular position of the object using a neural network model. The confidence value represents a likelihood that the object is an object of interest, and the neural network model is trained to generate lower confidence values for objects of the particular type when the objects of the particular type are occupying the particular position than when the objects of the particular type are not occupying the particular position.

Blind spot implementation in neural networks

In one respect, there is provided a system that may include a processor and a memory. The memory may be configured to store instructions that results in operations when executed by the processor. The operations may include: training a machine learning model by at least processing a training set with the machine learning model, the training set including at least one synthetic image that is generated by applying one or more modifications to a non-synthetic image; determining, based at least on a result of the processing of the mixed training set, that the machine learning model is unable to classify images having a specific modification; and training the machine learning model with additional training data that includes one or more additional synthetic images having the specific modification. Related methods and articles of manufacture are also disclosed.

Training machine learning models

In one respect, there is provided a system that may include a processor and a memory. The memory may be configured to store instructions that results in operations when executed by the processor. The operations may include: processing an image set with a convolutional neural network configured to detect, in the image set, a first feature and a second feature; determining a respective effectiveness of the first feature and the second feature in enabling the convolutional neural network to classify images in the image set; determining, based at least on the respective effectiveness of the first feature and the second feature, a first initial weight for the first feature and a second initial weight for the second feature; and initializing the convolutional neural network for training, the initialization of the convolutional neural network comprising configuring the convolutional neural network to apply the first initial weight and the second initial weight.

Weight initialization for machine learning models

In one respect, there is provided a system that may include a processor and a memory. The memory may be configured to store instructions that results in operations when executed by the processor. The operations may include: generating a concatenated feature map set by at least combining a first feature map set and a second feature map set, the first feature map set and the second feature map set each indicating one or more occurrences of a feature within an image, the first feature map set having a different scale than the second feature map set; and classifying, by a convolutional neural network, an image, the classifying of the image being based at least on the concatenated feature map set. Related methods and articles of manufacture are also disclosed.

Small object detection

A dataset of a plurality of transactions is received, the plurality of transactions including at least a first item and a second item. An absence-based association rule correlating an absence of the first item with an inclusion of the second item within the plurality of transactions is calculated.

Frequent item-set mining based on item absence

Methods, systems, and computer-readable storage media for receiving a set of frames, each frame being provided as a digital image that depicts a portion of an event and a logo associated with a brand, for each frame in the set of frames, and for each pixel in a frame: determining a weight of the pixel based on a distribution assigned to the frame, providing a quality of the logo depicted in the frame, and calculating a pixel quotient based on the weight and the quality, for each frame in the set of frames: determining a frame quotient at least partially based on a sum of all pixel quotients for the frame, and determining an impact indicator for the logo based on a total size of digital media comprising the set of frames and a sum of frame quotients of the frames in the set of frames.

Quantifying brand visual impact in digital media

The present disclosure describes methods, systems, and computer program products for improving frequent itemset mining. One computer-implemented method includes receiving, by operation of a computer system, a dataset of multiple transactions, each of the multiple transactions including one or more items; determining, by operation of the computer system, a ubiquitousness parameter of an item representing a frequency of occurrences of the item in the dataset; filtering, by operation of the computer system, out from the dataset one or more items that have ubiquitousness parameters larger than a ubiquitousness threshold; identifying, by operation of the computer system, one or more association rules that satisfy a confidence constraint based on the filtered dataset; and outputting, by operation of the computer system, the one or more association rules that satisfy the confidence constraint.

Enhancing frequent itemset mining

A number of parking segments is defined as a plurality of areas for street parking. A request for parking within the plurality of parking segments is received that defines parking constraints. The request is for optimize navigation to search for a free parking spot. Parking segments that comply with the parking constraints are determined. Current values for probabilities to park for the parking segments are determined based on evaluation of collected data for parking events. The data may include both traffic data and data related to defined features that are associated with determining of probability values. A route for traversing the parking segments is determined. The route defines an optimized order for traversing according to the determined current value for the probabilities to park. Navigation instructions are provided for the determined optimized route in response to the received request.

Optimized generation of navigation instructions based on computed parking probability values

A method and system for sharing trusted contact information between trusted, known and unknown parties, without revealing the contact information itself, thus protecting the party defined in the contact from possible unsolicited messages known as 'Spam'. Addresses of trusted senders are encrypted using a one-way encryption and stored in a whitelist repository that can be shared by different users. When a message is received, its senders address is extracted, encrypted using the same encryption method, and compared if it is found in the whitelist repository.

Method and system for sharing trusted contact information

An association method for analyzing a dataset and grouping its objects in homogeneous groups. Multiple grouping algorithms of the art are performed against the dataset, each producing its own grouping results. All the distributions (groupings) are ordered in a matrix, where each distribution suggested by a specific algorithm and/or parameters is presented in a column of the matrix, and each object is presented in a row of the matrix. The association method is applied to the grouping results and associates similar groups among different grouping methods. The best grouping of the objects is determined according to one or more quality criteria. The various grouping results are represented by colors in the matrix such that similar colors, even on different rows, represent similar grouping recommendations.

Method for efficient association of multiple distributions

A hierarchical collection of items including one or more sub collections of items ordered in a hierarchy is received. A statistical measure of frequency of an item in a sub collection of items is determined. Further, statistical measures of weightages of the item are determined defining a number of sub collections in the hierarchical collection of items in which the item appears and a number of sub collections in which the item appears. A statistical measure of variability defining a number of occurrences of the item in the hierarchical collection of items across different sub collections is calculated. Furthermore, a relevance score of the item is determined based on the statistical measure of frequency, the one or more statistical measures of weightages of the item and the statistical measure of variability. Based on the relevance score, the item is presented on a computer generated graphical user interface.

Growing community of inventors

Tel Aviv, Israel

Ran Bittmann