Abstract: The subject technology includes an image generator that generates images having specific visual concepts. The image generator uses a selective training process to fine-tune a text to a image generative system. The constrained text to image generative system may be trained to understand multiple custom tokens that embody visual characteristics of images included in fine-tuning datasets. Image generation prompts including one or more custom tokens may be used to condition the image creation process of the constrained text to image system to produce synthetic images having improved specificity, more creativity, and higher performance. Images created by the constrained text to image system may be ranked based on one or more criteria to further refine the created images for one or more specific use cases.

Abstract: The subject technology identifies a series of journey event types in an online user journey, the event types including an impression event, an email event, a click event, and a website visit, and assigns an encoder to each event type. Using an assigned encoder, the technology encodes each event type to generate an encoded vector for each event type. The encoded vector is representative of at least a portion of the online user journey relating to that event type. The technology generates an encoded vector for each event type to create a set of encoded vectors, the set of encoded vectors including one or more of an impression event encoded vector, an email event encoded vector, a click event encoded vector, and a website visit encoded vector.

Abstract: In some examples, a computerized system for analyzing images comprises at least one programmable processor and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising training an autoencoder using a plurality of image model training samples, the autoencoder comprising a plurality of interconnected layers and combined instances of neural networks, passing input data into a trained autoencoder model, the input data including at least one pixel image, encoding the input data into a compressed version of the input data, and decoding the compressed version of the input data to generate to create an output, the output including a sparse reconstruction of the input data, the output including a predicted pixel image label or score.

Abstract: In some examples, a system comprises at least one programmable processor; and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising: receiving a first request from at least one user device to execute an instance of an application; transmitting a graphical user interface (GUI) to the at least one user device to be rendered on a display of the at least one user device; receiving a second request, via the GUI, from the at least one user device, to deploy a digital advertisement, the second request including a set of platforms of a plurality of platforms of a multi-platform integration system, a set of settings, a set of parameters, and a set of allocation data; interfacing with each one of the platforms in the set of platforms; and integrating a digital advertisement directly with each one of the platforms in the set of platforms based on the set of settings, the

Abstract: A method comprises: collecting data including combinations of images and accompanying text and user feedback of the combinations; building training data sets based on the collected data; training a plurality of neural networks using the training data; generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks; generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks; generating a sequence of words based on the vectors using a third network of the plurality of trained neural networks; and transmitting the generated sequence of words and the specified image to the target audience over a network.

Abstract: The subject technology identifies a series of journey event types in an online user journey, the event types including an impression event, an email event, a click event, and a website visit, and assigns an encoder to each event type. Using an assigned encoder, the technology encodes each event type to generate an encoded vector for each event type. The encoded vector is representative of at least a portion of the online user journey relating to that event type. The technology generates an encoded vector for each event type to create a set of encoded vectors, the set of encoded vectors including one or more of an impression event encoded vector, an email event encoded vector, a click event encoded vector, and a website visit encoded vector.

Abstract: The subject technology uses a bootstrapping approach to train language models (LMs) to explain outcomes determined by neural networks, ensemble models, reinforcement learning models, LMs, and other black box machine learning models. The bootstrapping approach may train multiple iterations of a tuned LM using training data determined from progressively complex machine learning models and progressively detailed natural language explanations. The model explanations determined by the tuned LM may be displayed in a user interface (UI) included in a publishing system to provide users more context about and a greater understanding of the decision making process used by the black box machine learning models to determine outcomes.

Abstract: In some examples, a computerized system for analyzing images comprises at least one programmable processor and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising training an autoencoder using a plurality of image model training samples, the autoencoder comprising a plurality of interconnected layers and combined instances of neural networks, passing input data into a trained autoencoder model, the input data including at least one pixel image, encoding the input data into a compressed version of the input data, and decoding the compressed version of the input data to generate to create an output, the output including a sparse reconstruction of the input data, the output including a predicted pixel image label or score.

Abstract: In some examples, a system comprises at least one programmable processor; and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising: receiving a first request from at least one user device to execute an instance of an application; transmitting a graphical user interface (GUI) to the at least one user device to be rendered on a display of the at least one user device; receiving a second request, via the GUI, from the at least one user device, to deploy a digital advertisement, the second request including a set of platforms of a plurality of platforms of a multi-platform integration system, a set of settings, a set of parameters, and a set of allocation data; interfacing with each one of the platforms in the set of platforms; and integrating a digital advertisement directly with each one of the platforms in the set of platforms based on the set of settings, the

Abstract: The subject technology uses an agent architecture for language models and large language models (LMs) to complete a variety of different tasks within software platforms. The agent LMs are trained to determine different action chains that may be used to generate responses to tasks requested by users. The action chains may include a sequence of multiple actions that each complete a portion of the requested task. The agent LMs may be trained to perform different types of action chains using training prompts that teach the agent LMs to use tools that enable the LMs to interact with different software resources. The agent architecture may coordinate multiple agent LMs to complete tasks that require multiple action chains to complete.

Abstract: A method comprises: collecting data including combinations of images and accompanying text and user feedback of the combinations; building training data sets based on the collected data; training a plurality of neural networks using the training data; generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks; generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks; generating a sequence of words based on the vectors using a third network of the plurality of trained neural networks; and transmitting the generated sequence of words and the specified image to the target audience over a network.

Abstract: The subject technology optimizes media requests to improve the efficiency and reduce the costs of online media campaigns. The request optimization system may implement one or more ensemble learning techniques that leverage multiple machine learning systems trained on different datasets. The request optimization system may use the ensemble learning techniques to generate optimized media requests that account for one or more campaign goals and minimize price inefficiencies incurred while purchasing placements in online media exchanges. In various embodiments, dynamic data including real time exchange and impression data may be collected and used to retrain one or more machine learning systems. Retaining the machine learning systems on dynamic data may improve the performance of optimized media requests determined by the retrained systems.

Abstract: A system for training a bidding model comprising: a plurality of tactics stored on at least one database; a plurality of hyperparameters; in response to an available inventory from a publisher relayed through a real time bid server, computing a bid on the available inventory; sending the bid to the real time bid server; receiving an auction result in response to the bid; calculating a plurality of rewards based on the auction result and the tactics; calculate a plurality of q values based on the rewards; calculate a plurality of losses; backpropogating the losses through the bidding model.

Abstract: The subject technology detects anomalies in media campaign configuration settings. The anomaly detection system may leverage one or more deep learning models to detect anomalies and identify particular configuration settings that contribute to the detected anomalies. In various embodiments, two or more of the deep learning models may be combined into an ensemble model that boosts the accuracy of anomaly predictions made by the anomaly detection system. The anomaly detection system may review the configuration settings of media campaigns during the configuration process and before the media campaigns run on a publication system in order to reduce the amount of unsuccessful campaigns and minimize the amount of wasted resources spent on running campaigns that have a low likelihood of achieving user defined goals.

Abstract: In some examples, a system comprises at least one programmable processor; and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising: receiving a first request from at least one user device to execute an instance of an application; transmitting a graphical user interface (GUI) to the at least one user device to be rendered on a display of the at least one user device; receiving a second request, via the GUI, from the at least one user device, to deploy a digital advertisement, the second request including a set of platforms of a plurality of platforms of a multi-platform integration system, a set of settings, a set of parameters, and a set of allocation data; interfacing with each one of the platforms in the set of platforms; and integrating a digital advertisement directly with each one of the platforms in the set of platforms based on the set of settings, the

Abstract: The subject technology identifies a series of journey event types in an online user journey, the event types including an impression event, an email event, a click event, and a website visit, and assigns an encoder to each event type. Using an assigned encoder, the technology encodes each event type to generate an encoded vector for each event type. The encoded vector is representative of at least a portion of the online user journey relating to that event type. The technology generates an encoded vector for each event type to create a set of encoded vectors, the set of encoded vectors including one or more of an impression event encoded vector, an email event encoded vector, a click event encoded vector, and a website visit encoded vector.

Abstract: In some examples, a system comprises at least one programmable processor; and a machine-readable medium having instructions stored thereon which, when executed by the at least one programmable processor, cause the at least one programmable processor to execute operations comprising: receiving a first request from at least one user device to execute an instance of an application; transmitting a graphical user interface (GUI) to the at least one user device to be rendered on a display of the at least one user device; receiving a second request, via the GUI, from the at least one user device, to deploy a digital advertisement, the second request including a set of platforms of a plurality of platforms of a multi-platform integration system, a set of settings, a set of parameters, and a set of allocation data; interfacing with each one of the platforms in the set of platforms; and integrating a digital advertisement directly with each one of the platforms in the set of platforms based on the set of settings, the

Abstract: The subject technology identifies a series of journey event types in an online user journey, the event types including an impression event, an email event, a click event, and a website visit, and assigns an encoder to each event type. Using an assigned encoder, the technology encodes each event type to generate an encoded vector for each event type. The encoded vector is representative of at least a portion of the online user journey relating to that event type. The technology generates an encoded vector for each event type to create a set of encoded vectors, the set of encoded vectors including one or more of an impression event encoded vector, an email event encoded vector, a click event encoded vector, and a website visit encoded vector.

Abstract: A system for training a bidding model comprising: a plurality of tactics stored on at least one database; a plurality of hyperparameters; in response to an available inventory from a publisher relayed through a real time bid server, computing a bid on the available inventory; sending the bid to the real time bid server; receiving an auction result in response to the bid; calculating a plurality of rewards based on the auction result and the tactics; calculate a plurality of q values based on the rewards; calculate a plurality of losses; backpropogating the losses through the bidding model.

Abstract: A system for training a bidding model comprising: a plurality of tactics stored on at least one database; a plurality of hyperparameters; in response to an available inventory from a publisher relayed through a real time bid server, computing a bid on the available inventory; sending the bid to the real time bid server; receiving an auction result in response to the bid; calculating a plurality of rewards based on the auction result and the tactics; calculate a plurality of q values based on the rewards; calculate a plurality of losses; backpropogating the losses through the bidding model.