Machine Learning is becoming more and more important in research and in daily life, yet the learning process remains largely misunderstood and many important questions are still unresolved. Statistical physicists, in a long tradition of looking for universal behavior and simple mechanisms to understand complex collective phenomena, have taken their turn in trying to understand these models with their own language. It is therefore natural that the Boltzmann Machine - or the inverse Ising problem- enters at the crossroad of statistical physics and machine learning
Machine Learning is becoming more and more important in research and in daily life, yet the learning process remains largely misunderstood and many important questions are still unresolved. Statistical physicists, in a long tradition of looking for universal behavior and simple mechanisms to understand complex collective phenomena, have taken their turn in trying to understand these models with their own language. It is therefore natural that the Boltzmann Machine - or the inverse Ising problem- enters at the crossroad of statistical physics and machine learning
Conventionally, the study of phases in statistical mechan- ics is performed with the help of random sampling tools. Among the most powerful are Monte Carlo simulations consisting of a stochastic importance sampling over state space and evaluation of estimators for physical quantities. The ability of modern machine learning techniques to classify, identify, or in- terpret massive data sets provides a complementary paradigm to the above approach to analyze the exponentially large number of states in statistical physics. In this report, it is demonstrated by application on Ising-type models that deep learning has potential wide applications in solving many-body statis- tical physics problems. In application of supervised learning, we showed that the feed-forward neural network can identify phases and phase transitions in the ferromagnetic Ising model and the convolutional neural network (CNN) is extremely powerful in classifying T = 0 and T = ∞ phases in the Ising gauge model; In application of unsupervised learning, we illustrated that a deep auto-encoder constructed by stacked restricted Boltzmann machines (RBM)\nis closely related to the renormalization group (RG) method well understood in modern physics and our reconstruction of Ising spin configurations in the ferromagnetic Ising model is similar to the hand-written digits reconstruction.
Questo libro fornisce un'introduzione completa al campo delle transizioni di fase che permeano molte aree della fisica. I concetti alla base di questi fenomeni sono parte del linguaggio moderno della fisica dello stato solido e della meccanica statistica, della fisica delle particelle elementari così come della recente teoria delle stringhe. Il modello di Ising, tra i più noti della meccanica statistica, è una guida per avvicinarsi alla teoria generale dei fenomeni critici poiché la sua qualità essenziale, la semplicità, consente di evidenziarne facilmente gli aspetti fondamentali. Tuttavia, la sua ricchezza teorica e matematica costituisce il punto di partenza per l'approfondimento di molti altri temi ormai entrati nel bagaglio scientifico dei moderni ricercatori in fisica e in matematica. Il testo affronta gli argomenti trattati con un approccio didattico, includendo una larga parte di materiale di base in meccanica statistica, teoria dei campi, fisica dello stato solido, analisi complessa, algebra di Lie. Ciascun capitolo è accompagnato da una serie di esercizi di ampio spettro con diversi gradi di difficoltà. Testo utile per i vari corsi di dottorato in Fisica, questo libro è anche rivolto agli studenti dei corsi di laurea specialistica in Fisica e in Matematica e, in generale, a tutti coloro che sono interessati ai recenti sviluppi di quest'area molto viva della ricerca contemporanea.
Machine learning has gained in popularity over recent years.It shows great power in many real-world problems, including Computer Vision, Natural Language Processing and Speech Recognition.However, a clear theoretical mechanism is still lacked behind many machine learning algorithms, such as Deep Learning.Good theoretical analysis can not only explain the current success of machine learning, but also reveal new approaches to future advancements.In last few years, more and more physicists contributed good works that provided novel insights and philosophy for machine learning communities.Due to two reasons, we believe physical approaches can be very promising.First, we think machine learning share some similar theoretical mechanism with physical theories, particularly quantum physics and statistical physics.We may employ physical languages to study machine learning models in some physical frameworks.Second, real-world problems and information itself must obey some fundamental laws, which are often physical laws for our universe.Good physical prior knowledge which may be helpful to reduce model complexity and improve generalization.The physical perspective deserve more attention.In this thesis, we review a few related works on the physics based approach and related traditional machine learning.It helps us understand how physics may interact with machine learning.We propose quantum interpretations for machine learning and a class of original physics-inspired machine learning algorithms, named Quantum-Inspired Forest.Both theoretical proof and empirical analysis are presented in details.The success of quantum-inspired machine learning encourages us to pay attention to multiple physical viewpoints.Even without solid theoretical analysis, important prior knowledge and heuristics may be obtained from physical models.The proposed algorithm is an example that illustrates how to combine physics and machine learning together.We in particular discuss a physical perspective on machine learning.We hope our physical analysis on machine learning would inspire more novel works in near future.
In recent decades, machine learning has emerged as a very powerful computational method. Because of its exceptional successes in computer science and engineering, machine learning has ignited research interest in other disciplines, including biology, chemistry, physics, and finance. Machine learning models, which are usually regarded as mathematical models, have traditionally been implemented on the basis of digital computing platform (Figure 1A). The increasing prevalence of machine learning has been accompanied by a rapid increase of computing requirements, outpacing Moore’s law.
Neural network approaches to machine learning problems rely on statistical frameworks that are designed to generalize to many situations, whereas traditional physics-based approaches to quantum problems stem from specific physical laws that govern the behavior of certain classes of systems. While quantum physics and machine learning have been largely disparate fields for most of their history, their interplay has gained considerable attention in recent years, primarily driven by the premise that the most powerful computational models are those whose foundations stem from both physical and statistical principles. Two overarching strategies in this area include building physics-based machine learning models from the ground up, or alternatively, looking for hidden ways that statistical models include aspects of physical systems. Many-body quantum systems turn out to be a useful testing ground for both of these approaches, due to the complex behavior that arises from simple fundamental rules. Here, we present several ways that neural networks can represent many-body spin states, which enables the exploration into the underlying nature of neural networks, as well as gaining insight into how effective different classes of models are for simulating physical systems. The main focus areas of this thesis are (1) connecting convolutional neural networks to maximum-entropy distributions when applied to spin systems, (2) using generative models to gain physical insights into experimental tomographic data, and (3) applying inverse reinforcement learning to probe the nature of agent-based quantum circuit design.