Deep learning and data analysis. Practical guidance

I read this one. book Dmitry Malov for a couple of weeks that fell on vacation and travel. The book is published in 2023 and is accompanied by a code and schedule that accessible Github. In the book 270 pages divided into 8 The author tries to first set out the necessary foundations, and then proceed to practice and examples: 1. Fundamentals of machine learning The author begins with the basis of the Linear algebra - scalar, vector, matrix, tensor, norm Information theory and theorver - random value, probability distribution, conditional probability, expectation, variance, covariance, Bayes' rule

• Basic concepts of machine learning and solved problems - classification, regression, anomaly detection, machine translation, structural inference, synthesis and sampling), as well as attitudes to learning experience and learning options with a teacher, without a teacher, with partial involvement of a teacher, as well as reinforcement learning Basics of development: python syntax, basics of oop (Abstraction, encapsulation, polymorphism, inheritance and composition)Development processes: waterfall and agile:) 2. The main algorithms of machine learning - here the author begins with the preprocessing of data, and then considers the algorithms for reducing their dimensionality: linear and nonlinear methods, linear and logistic regression, decision trees, the reference vector method, the naive Bayesian classifier, k-means, nearest k neighbors, random forest and gradient boosting algorithm. It all fits in. 30 pages, so if you have not separately studied all these things, then sometimes it is difficult to keep up with the thought of the author. (Thank goodness it's been going on for years. 10 back when I had a burst of self-education and was hanging out on Coursera and Edx) 3. Fundamentals of Deep Learning The author begins by propagating the error. (backpropagation), further talks about the perceptron, the Markov chain, the Boltzmann machine, the Hopfield network, convolutional neural networks (CNN)Transformers, Recurrent Neural Networks (RNN)Autocoders, generatively competitive networks (GAN). And at the end, he gives an example of a system that the author did for a whitepaper or diploma:) 4. Basics of data science It is interesting that here the story begins with the methodology of working with data, or rather with data. CRISP-DM (Cross-Industry Standard Process for Data Mining, further describes the roles in the ML-development team, where data analyst, data engineer, data scientist are notable. Next, the author talks about the trends: deep fakes and combating them, business interest in training end2end models, Auto ML for low-code and no-code use, MLOps (recently wrote About the whitepaper from Google and participated podcast). And the author concludes this chapter by considering popular libraries for ML development, among which I would like to mention TensorFlow, PyTorch, Keras. 5. The challenges of deep learning In this chapter, the author gives examples of problems and shows how they can be solved using deep learning. I'm gonna need a code here. repositoryTo play with the tasks yourself. There are examples of data augmentation, computer vision and the use of OpenCV, the classic problem of character recognition, natural language processing, audio processing, and video processing. In general, in this and the next three chapters, the pulp is collected:) 6-8. The last three chapters are about getting to know TensorFlow, Keras and PyTorch.. It shows how to solve problems from 5 chapters using a specific library.

If I finalize the sammari book, it seemed to me a brief intro to deep learning. It has all the necessary basic information, but to understand it you will have to read additional materials. Here are examples of tasks and code that can be the starting point for your experiments. I liked the book, but except for the library. TensorFlow, Keras and PyTorch I’ve studied the rest a long time ago and it’s pretty good:)

#AI #Math #Statistics #Software #DataScience #ML