Estimating the entropy of a discrete random variable is a fundamental problem in information theory and related fields. This problem has many applications in various domains, including machine learning, statistics, and data compression. Over the years, a variety of estimation schemes have been suggested. However, despite significant progress, most methods still struggle when the sample is small, compared to the variable’s alphabet size. In this work, we introduce a practical solution to this problem, which extends the work of McAllester and Statos. The proposed scheme uses the generalization abilities of cross-entropy estimation in deep neural networks (DNNs) to introduce improved entropy estimation accuracy. Furthermore, we introduce a family of estimators for related information-theoretic measures, such as conditional entropy and mutual information (MI). We show that these estimators are strongly consistent and demonstrate their performance in a variety of use cases. First, we consider large alphabet entropy estimation. Then, we extend the scope to MI estimation. Next, we apply the proposed scheme to conditional MI estimation, as we focus on independence testing tasks. Finally, we study a transfer entropy (TE) estimation problem. The proposed estimators demonstrate improved performance compared to existing methods in all of these setups.

We propose a new algorithm called the context-based predictive information (CBPI) for estimating the predictive information (PI) between time series, by utilizing a lossy compression algorithm. The advantage of this approach over existing methods resides in the case of sparse predictive information (SPI) conditions, where the ratio between the number of informative sequences to uninformative sequences is small. It is shown that the CBPI achieves a better PI estimation than benchmark methods by ignoring uninformative sequences while improving explainability by identifying the informative sequences. We also provide an implementation of the CBPI algorithm on a real dataset of large banks’ stock prices in the U.S. In the last part of this paper, we show how the CBPI algorithm is related to the well-known information bottleneck in its deterministic version.