Insomework,functionapproximation scheme is adopted such that essential quantities for policy improvement, e.g.

Motivated by the prevailing paradigm of using unsupervised learning for efficient exploration in reinforcement learning (RL) problems [tang2017exploration,bellemare2016unifying], we investigate when this paradigm is provably efficient. We study episodic Markov decision processes with rich observations generated from a small number of latent states. We present a general algorithmic framework that is built upon two components: an unsupervised learning algorithm and a no-regret tabular RL algorithm. Theoretically, we prove that as long as the unsupervised learning algorithm enjoys a polynomial sample complexity guarantee, we can find a near-optimal policy with sample complexity polynomial in the number of latent states, which is significantly smaller than the number of observations. Empirically, we instantiate our framework on a class of hard exploration problems to demonstrate the practicality of our theory.

Deep learning models struggle with compositional generalization, i.e. the ability to recognize or generate novel combinations of observed elementary concepts. In hopes of enabling compositional generalization, various unsupervised learning algorithms have been proposed with inductive biases that aim to induce compositional structure in learned representations (e.g.

In this work, we evaluate these unsupervised learning algorithms in terms of how well they enable compositional generalization .

Motivated by the prevailing paradigm of using unsupervised learning for efficient exploration in reinforcement learning (RL) problems [tang2017exploration,bellemare2016unifying], we investigate when this paradigm is provably efficient. We study episodic Markov decision processes with rich observations generated from a small number of latent states. We present a general algorithmic framework that is built upon two components: an unsupervised learning algorithm and a no-regret tabular RL algorithm. Theoretically, we prove that as long as the unsupervised learning algorithm enjoys a polynomial sample complexity guarantee, we can find a near-optimal policy with sample complexity polynomial in the number of latent states, which is significantly smaller than the number of observations. Empirically, we instantiate our framework on a class of hard exploration problems to demonstrate the practicality of our theory.

Deep learning models struggle with compositional generalization, i.e. the ability to recognize or generate novel combinations of observed elementary concepts. In hopes of enabling compositional generalization, various unsupervised learning algorithms have been proposed with inductive biases that aim to induce compositional structure in learned representations (e.g. In this work, we evaluate these unsupervised learning algorithms in terms of how well they enable \textit{compositional generalization}. Specifically, our evaluation protocol focuses on whether or not it is easy to train a simple model on top of the learned representation that generalizes to new combinations of compositional factors. We systematically study three unsupervised representation learning algorithms - \beta -VAE, \beta -TCVAE, and emergent language (EL) autoencoders - on two datasets that allow directly testing compositional generalization.

The constant increase of devices connected to the Internet, and therefore of cyber-attacks, makes it necessary to analyze network traffic in order to recognize malicious activity. Traditional packet-based analysis methods are insufficient because in large networks the amount of traffic is so high that it is unfeasible to review all communications. For this reason, flows is a suitable approach for this situation, which in future 5G networks will have to be used, as the number of packets will increase dramatically. If this is also combined with unsupervised learning models, it can detect new threats for which it has not been trained. This paper presents a systematic review of the literature on unsupervised learning algorithms for detecting anomalies in network flows, following the PRISMA guideline. A total of 63 scientific articles have been reviewed, analyzing 13 of them in depth. The results obtained show that autoencoder is the most used option, followed by SVM, ALAD, or SOM. On the other hand, all the datasets used for anomaly detection have been collected, including some specialised in IoT or with real data collected from honeypots.

Deep learning models struggle with compositional generalization, i.e. the ability to recognize or generate novel combinations of observed elementary concepts. In hopes of enabling compositional generalization, various unsupervised learning algorithms have been proposed with inductive biases that aim to induce compositional structure in learned representations (e.g. In this work, we evaluate these unsupervised learning algorithms in terms of how well they enable \textit{compositional generalization}. Specifically, our evaluation protocol focuses on whether or not it is easy to train a simple model on top of the learned representation that generalizes to new combinations of compositional factors. We systematically study three unsupervised representation learning algorithms - \beta -VAE, \beta -TCVAE, and emergent language (EL) autoencoders - on two datasets that allow directly testing compositional generalization.