With case research, Python code, new open supply libraries, and purposes of the  GenAI game-changer expertise referred to as NoGAN (194 pages). 

This e book covers optimization methods pertaining to machine studying and generative AI, with an emphasis on producing higher artificial information with sooner strategies, some not even involving neural networks. NoGAN for tabular information is described intimately, together with full Python code, and case research in healthcare, insurance coverage, cybersecurity, schooling, and telecom. This low-cost approach is a sport changer: it runs 1000x sooner than generative adversarial networks (GAN) whereas persistently producing higher outcomes. Additionally, it results in replicable outcomes and auto-tuning.

Many analysis metrics fail to detect defects in synthesized information, not as a result of they’re unhealthy, however as a result of they’re poorly applied: as a result of complexity, the complete multivariate model is absent from vendor options. On this e book, I describe an implementation of the complete model, examined on quite a few examples. Often called the multivariate Kolmogorov-Smirnov distance (KS), it’s based mostly on the joint empirical distributions connected to the datasets, and work in any dimension on categorical and numerical options. Python libraries, each for NoGAN and KS, at the moment are accessible and offered on this e book.

A really completely different synthesizer additionally mentioned, particularly NoGAN2, is predicated on resampling, model-free hierarchical strategies, auto-tuning, and explainable AI. It minimizes a specific loss perform, additionally with out gradient descent. Whereas not based mostly on neural networks, it nonetheless shares many similarities with GAN. Thus you need to use it as a sandbox to shortly take a look at varied options and hyperparameters earlier than including those that work finest, to GAN. Although NoGAN and NoGAN2 don’t use conventional optimization, gradient descent is the subject of the primary chapter. Utilized to information fairly than math capabilities, there is no such thing as a assumption of differentiability, no studying parameter, and primarily no math. The second chapter introduces a generic class of regression strategies overlaying all present ones and extra, whether or not your information has a response or not, for supervised or unsupervised studying. I take advantage of gradient descent on this case.

One chapter is dedicated to NLP, that includes an environment friendly approach to course of massive quantities of textual content information: hidden resolution timber, presenting some similarities with XGBoost. An analogous approach is utilized in NoGAN. Then I talk about different GenAI strategies and varied optimization methods, together with function clustering, information thinning, good grid search and extra. Multivariate interpolation is used for time sequence and geospatial information, whereas agent-based modeling applies to advanced techniques.

Strategies are accompanied by enterprise-grade Python code, additionally accessible on GitHub. Chapters are largely impartial from one another, permitting you to learn in random order. The model could be very compact, and appropriate to enterprise professionals with little time. Jargon and arcane theories are absent, changed by easy English to facilitate the studying by non-experts, and that will help you uncover matters normally made inaccessible to rookies. Whereas state-of-the-art analysis is current in all chapters, the stipulations to learn this e book are minimal: an analytic skilled background, or a primary course in calculus and linear algebra.

Revealed October 2023. See desk of contents right here. You should buy the e book right here.

Writer

Vincent Granville is a pioneering GenAI scientist and machine studying skilled, co-founder of Knowledge Science Central (acquired by a publicly traded firm in 2020), Chief AI Scientist at MLTechniques.com, former VC-funded government, writer and patent proprietor — one associated to LLM. Vincent’s previous company expertise consists of Visa, Wells Fargo, eBay, NBC, Microsoft, and CNET.

Vincent can also be a former post-doc at Cambridge College, and the Nationwide Institute of Statistical Sciences (NISS). He revealed in Journal of Quantity Concept,  Journal of the Royal Statistical Society (Collection B), and IEEE Transactions on Sample Evaluation and Machine Intelligence. He’s the writer of a number of books, together with “Artificial Knowledge and Generative AI” (Elsevier, 2024). Vincent lives in Washington state, and enjoys doing analysis on stochastic processes, dynamical techniques, experimental math and probabilistic quantity concept. He just lately launched a GenAI certification program, providing state-of-the-art, enterprise grade tasks to members.