Efficient Likelihood Function Reparametrization for Regression against Categorical Variables

We simplify the likelihood function obtained when regressing on categorical variables. This speeds up the variational inference implementation in tensorflow.

Bayesian Statistics Tensorflow Probability

Monitor deployed tensorflow models with Prometheus and Grafana

We provide a minimal example how to serve tensorflow models on a Kubernetes cluster and monitor them with Prometheus and Grafana. To expose the models we create a deployment and service manifests, whereas for the deployment of Prometheus and Grafana we use the helm charts provided by bi...

Kubernetes TensorFlow Prometheus Grafana

MLflow on Kubernetes

We will containerize and deploy a MLflow server on a Kubernetes cluster on Google cloud. We will also create the MLflow backend DB, the artifact store and all required service accounts, and secrets on Google cloud. This is achieved by using either gcloud SDK or terraform. The deployment...

MLflow Kubernetes Google Cloud

Bayesian inference for stochastic processes: an analytically solvable problem

We explain the application of the Bayesian inference approach to the case of having multiple trajectories of a stochastic process. We will consider an analytically solvable problem to address the question of how much the past values of a trajectory reduce our uncertainty about its futur...

Time Series Analysis Bayesian Modeling

Variational inference in probabilistic models: an analytically solvable example

The Bayesian inference approach gives us the opportunity to systematically combine and update our prior beliefs about the model parameters with new evidence. In the case where the prior and posterior are conjugate distributions, we can find either an exact analytic or a numerically inex...

Variational Inference Bayesian Modeling

Geospatial data visualization

Visualize random locations, vehicle trajectories and vehicle telematics data.

React Geospatial Visualization

Ad auction bidding strategy

Real-Time Bidding (RTB) has become a relevant paradigm in display advertising. It mimics stock exchanges and utilizes computer algorithms to buy and sell ads in real-time automatically. Imagine that you have to participate in $N >> 1$ of those online ad auctions with a limited bidding b...

Bidding Optimization

Track the model performance metrics in Federated training

Federated learning is a machine learning technique that trains a model across multiple decentralized devices, each of them holding a local data sample, without exchanging these data samples. Let’s imagine that by using this technique you have trained a binary classification model. You w...

Federated Learning

Navigation through a dynamic map using the Bellman equation

We will look at the problem of navigating through a dynamically changing map. It can be represented as a sequence of optimization problems for every time step and in the end, it will reduce to a specific case of the Bellman equation. A solution will be discussed and applied to a particu...

Reinforcement Learning Bellmann Equation