Machine learning can identify phases and phase transitions in condensed matter systems through supervised learning. A standard feed-forward neural network can be trained to detect multiple types of order parameters from raw Monte Carlo samples. Convolutional neural networks can detect topological phases without conventional order parameters. These results show that machine learning can classify phases without knowledge of the Hamiltonian or interaction locality. Condensed matter physics studies complex systems with exponentially growing state spaces. Machine learning techniques can recognize and classify complex data. Modern machine learning, such as neural networks, can identify phases and phase transitions in condensed matter systems. Training neural networks on Monte Carlo data provides a powerful framework for supervised learning of phases and phase boundaries. Conventional methods use Monte Carlo simulations and order parameters to identify phases. However, some states may not be easily identified with standard estimators. Machine learning provides an alternative approach, capable of classifying data without prior knowledge of the Hamiltonian or interactions. The paper demonstrates the use of machine learning on the Ising model, showing that neural networks can classify states with high accuracy. The neural network learns representations of low- and high-temperature states, achieving 94% accuracy for N = 100 spins. The model generalizes to other systems, such as the triangular Ising model, where it estimates the critical temperature with high accuracy. The paper also explores the application of machine learning to disordered and topological phases, such as Coulomb phases and Ising lattice gauge theories. Neural networks can detect subtle differences in higher-order correlation functions. Convolutional neural networks are used to classify high- and low-temperature states in the Ising gauge theory, achieving 100% accuracy. These results show that neural networks can encode information about phases and detect phase transitions in complex systems. They can represent ground states with topological order, such as the toric code. Machine learning has the potential to become a fundamental research tool in condensed matter and statistical physics, with applications in quantum technologies and experiments with high-resolution quantum gas microscopes.Machine learning can identify phases and phase transitions in condensed matter systems through supervised learning. A standard feed-forward neural network can be trained to detect multiple types of order parameters from raw Monte Carlo samples. Convolutional neural networks can detect topological phases without conventional order parameters. These results show that machine learning can classify phases without knowledge of the Hamiltonian or interaction locality. Condensed matter physics studies complex systems with exponentially growing state spaces. Machine learning techniques can recognize and classify complex data. Modern machine learning, such as neural networks, can identify phases and phase transitions in condensed matter systems. Training neural networks on Monte Carlo data provides a powerful framework for supervised learning of phases and phase boundaries. Conventional methods use Monte Carlo simulations and order parameters to identify phases. However, some states may not be easily identified with standard estimators. Machine learning provides an alternative approach, capable of classifying data without prior knowledge of the Hamiltonian or interactions. The paper demonstrates the use of machine learning on the Ising model, showing that neural networks can classify states with high accuracy. The neural network learns representations of low- and high-temperature states, achieving 94% accuracy for N = 100 spins. The model generalizes to other systems, such as the triangular Ising model, where it estimates the critical temperature with high accuracy. The paper also explores the application of machine learning to disordered and topological phases, such as Coulomb phases and Ising lattice gauge theories. Neural networks can detect subtle differences in higher-order correlation functions. Convolutional neural networks are used to classify high- and low-temperature states in the Ising gauge theory, achieving 100% accuracy. These results show that neural networks can encode information about phases and detect phase transitions in complex systems. They can represent ground states with topological order, such as the toric code. Machine learning has the potential to become a fundamental research tool in condensed matter and statistical physics, with applications in quantum technologies and experiments with high-resolution quantum gas microscopes.