Utilizing a deep generative model with task embedding for personalized targeting of digital content through multiple channels across client devices

A non-transitory computer readable medium comprising instructions that, when executed by at least one processor, cause a computing device to:

• identify a client device and digital characteristics corresponding to the client device;
• apply a trained generative machine learning model with a task embedding layer to the digital characteristics corresponding to the client device to generate a multi-treatment effect estimation of a first combination of treatments from a set of treatments by:
• utilizing the task embedding layer of the trained generative machine learning model to generate a first task embedding by processing the first combination of treatments from the set of treatments;
• analyzing the first task embedding of the first combination of treatments and the digital characteristics corresponding to the client device via the trained generative machine learning model to generate a first multi-treatment effect estimation of the first combination of treatments;
• and
• based on the first multi-treatment effect estimation of the first combination of treatments, select one or more treatments from the set of treatments to provide to the client device, the one or more treatments comprising one or more distribution channels for providing digital content to the client device.

A system comprising:

• at least one processor;
• a memory comprising:
• a variational autoencoder having a task embedding layer;
• a digital repository of historical training data comprising a training characteristic set corresponding to training client devices, training treatments, and training effects;
• and
• instructions that, when executed by the at least one processor, cause the system to train the variational autoencoder by:
• applying the variational autoencoder to a training characteristic set of the training characteristic sets to generate a predicted distribution of treatments in light of the training characteristic set;
• sampling the predicted distribution of treatments to generate a treatment vector representing a combination of treatments;
• utilizing the task embedding layer of the variational autoencoder to generate a task embedding of the treatments based on the treatment vector;
• analyzing the task embedding and the training characteristic set via the variational autoencoder to generate a predicted multi-treatment effect;
• and
• modifying parameters of the variational autoencoder by comparing the predicted multi-treatment effect with a training effect from the training effects.