Python Deep Learning Projects

9 projects demystifying neural network and deep learning models for building intelligent systems

Matthew Lamons , Rahul Kumar , Abhishek Nagaraja 지음

Packt(GCO Science) 2018년 10월 31일 출간

작품소개

책소개

▶Book Description
Deep learning has been gradually revolutionizing every field of artificial intelligence, making application development easier.

Python Deep Learning Projects imparts all the knowledge needed to implement complex deep learning projects in the field of computational linguistics and computer vision. Each of these projects is unique, helping you progressively master the subject. You’ll learn how to implement a text classifier system using a recurrent neural network (RNN) model and optimize it to understand the shortcomings you might experience while implementing a simple deep learning system.

Similarly, you’ll discover how to develop various projects, including word vector representation, open domain question answering, and building chatbots using seq-to-seq models and language modeling. In addition to this, you’ll cover advanced concepts, such as regularization, gradient clipping, gradient normalization, and bidirectional RNNs, through a series of engaging projects.

By the end of this book, you will have gained knowledge to develop your own deep learning systems in a straightforward way and in an efficient way

▶What You Will Learn
？ Set up a deep learning development environment on Amazon Web Services (AWS)
？ Apply GPU-powered instances as well as the deep learning AMI
？ Implement seq-to-seq networks for modeling natural language processing (NLP)
？ Develop an end-to-end speech recognition system
？ Build a system for pixel-wise semantic labeling of an image
？ Create a system that generates images and their regions

▶Key Features
？ Explore deep learning across computer vision, natural language processing (NLP), and image processing
？ Discover best practices for the training of deep neural networks and their deployment
？ Access popular deep learning models as well as widely used neural network architectures

▶Who This Book Is For
This book is perfect for you if you've undertaken at least one course in machine learning and have a modest functional proficiency in Python (meaning you can create programs in Python when supported by examples). Many of our readers will be undergraduates at university studying computer science, statistics, mathematics, physics, biology, chemistry, marketing, and business. Deep learning technologies are being applied to all the professions that these degrees prepare you for, and this book is a great way to learn skills that will be applicable to your success. Postgraduates will appreciate the instruction level, too, as the projects selected are directly applicable to the modern job market, from tech start-ups to enterprise applications.

Python Deep Learning Projects is focused at the core of the data science pipeline ？ model building, training, evaluation, and validation. Additional pre- and post-data science engineering processes are required in the data pipeline for production applications that we cannot address here due to space considerations, but that we are looking to address in a future publication.

목차

▶TABLE of CONTENTS
1: BUILDING DEEP LEARNING ENVIRONMENTS
2: TRAINING NN FOR PREDICTION USING REGRESSION
3: WORD REPRESENTATION USING WORD2VEC
4: BUILDING AN NLP PIPELINE FOR BUILDING CHATBOTS
5: SEQUENCE-TO-SEQUENCE MODELS FOR BUILDING CHATBOTS
6: GENERATIVE LANGUAGE MODEL FOR CONTENT CREATION
7: BUILDING SPEECH RECOGNITION WITH DEEPSPEECH2
8: HANDWRITTEN DIGITS CLASSIFICATION USING CONVNETS
9: OBJECT DETECTION USING OPENCV AND TENSORFLOW
10: BUILDING FACE RECOGNITION USING FACENET
11: AUTOMATED IMAGE CAPTIONING
12: POSE ESTIMATION ON 3D MODELS USING CONVNETS
13: IMAGE TRANSLATION USING GANS FOR STYLE TRANSFER
14: DEVELOP AN AUTONOMOUS AGENT WITH DEEP R LEARNING
15: SUMMARY AND NEXT STEPS IN YOUR DEEP LEARNING CAREER

책 속으로

▶What this book covers
？ Chapter 1, Building Deep Learning Environments, in this chapter we will establish a common workspace for our projects with core technologies such as Ubuntu, Anaconda, Python, TensorFlow, Keras, and Google Cloud Platform (GCP).

？ Chapter 2, Training a Neural Net for Prediction Using Regression, in this chapter we will build a 2 layer (minimally deep) neural network in TensorFlow and train it on the classic MNIST dataset of handwritten digits for a restaurant patron text notification business use case.

？ Chapter 3, Word Representations Using word2vec, in this chapter we will learn and use word2vec to transform words into dense vectors (that is, tensors) creating embedding representations for a corpus, then create a convolutional neural network (CNN) to build a language model for sentiment analysis in a text exchange business use case.

？ Chapter 4, Build a NLP Pipeline for Building Chatbots, in this chapter we will create an NLP pipeline to tokenized a corpus, tag parts of speech, determine relationships between words with dependency parsing, and that conducts Named Entity Recognition. Use TF-IDF to vectorize the features in the document to create a simple FAQ type chatbot. Enhance this chatbot with NER and implementation of Rasa NLU to build a chatbot which understands the context (intent) to provide an accurate response.

？ Chapter 5, Sequence-to-sequence Models for Building Chatbots, in this chapter we will use Chapter 4, Build a NLP Pipeline for Building Chatbots, chatbots to build a more advanced chatbot combining learnings from earlier projects in a number of technologies to make a chatbot that is more contextually aware and robust. We avoided some of the limitations of CNNs in chatbots by building a recurrent neural network (RNN) model with long short-term memory (LSTM) units specifically designed to capture the signal represented in sequences of characters or words.

？ Chapter 6, Generative Language Model for Content Creation, in this chapter we implement a generative model to generate content using the long short-term memory (LSTM), variational autoencoders, and Generative Adversarial Networks (GANs). You will effectively implement models for both text and music which can generate song lyrics, scripts, and music for artists and various creative businesses.

？ Chapter 7, Building Speech Recognition with DeepSpeech2, in this chapter we build and train an automatic speech recognition (ASR) system to accept then convert an audio call to text that could then be used as the input into a text-based chatbot. Work with speech and spectrograms to build an end-to-end speech recognition system with a Connectionist Temporal Classification (CTC) loss function, batch normalization and SortaGrad for the RNNs. This chapter is the capstone in the Natural Language Processing section of the Python Deep Learning Projects Book.

？ Chapter 8, Handwritten Digit Classification Using ConvNets, in this chapter we will teach the fundamentals of Convolutional Neural Networks (ConvNets) in an examination of the convolution operation, pooling, and dropout regularization. These are the levers you'll adjust in tuning your models in your career. See the value of deploying a more complex and deeper model in the performance results compared to an earlier Python Deep Learning Project in Chapter 2, Training a Neural Net for Prediction Using Regression.

？ Chapter 9, Object Detection Using OpenCV and TensorFlow, in this chapter we will learn to master object detection and classification while using significantly more informationally complex data than previous projects, to produce impressive outcomes. Learn to use the deep learning package YOLOv2 and gain experience how this model architecture gets deeper and more complex and produces good results.
...

출판사 서평

▶ Preface
Have you ever tried to get something novel out of a computer? I can ask you to make up a story or look at a picture and tell me what's in it. How would you make a computer program behave like this in contrast to the digital storage and transfer unit we've used them for these past 30+ yrs?

If you had perfect knowledge and all the time in the world, you could write all the rules by which a computer program would need to operate. Of course, if you had all the knowledge to define the operational rules, you wouldn't need the computer to do anything! So what do you do if you need a computer to function in complex ways (making predictions, classifications, optimizing processes, generating content, responding to interactions, performing robotic controls), but don't have all the heuristic rules defined?

You build an algorithmically-based application that can learn the rules, find the pattern, or determine the signal, from data that comes from the domain space in question. You set up the training such that it iterates incredibly fast and with a great number of cycles (we call them epochs) to provide the "experience" to incrementally train the model in a process that would not be possible in a human lifetime.

When we build these algorithmic architectures in layers, we create deep learning models that can learn features (for example, dogs have tails and cars have wheels) and these learned features are powerful! What we really find in Python Deep Learning Projects is that we can ask profound questions not possible before. It's these questions that drive our deep learning technologies to solve problems that range from healthcare diagnostics in radiology to cancer screening. Deep learning applications drive chatbot experiences, facial recognition, autonomous vehicles, recommendation engines, and marketing tech. The hard sciences of physics, biology, and chemistry are incorporating deep learning skills training just as they have in the past with regard to statistics and microscopes.

작가정보

저자(글) Matthew Lamons

Matthew Lamons's background is in experimental psychology and deep learning. Founder and CEO of Skejul―the AI platform to help people manage their activities. Named by Gartner, Inc. as a ""Cool Vendor"" in the ""Cool Vendors in Unified Communication, 2017"" report. He founded The Intelligence Factory to build AI strategy, solutions, insights, and talent for enterprise clients and incubate AI tech startups based on the success of his Applied AI MasterMinds group. Matthew's global community of more than 85 K are leaders in AI, forecasting, robotics, autonomous vehicles, marketing tech, NLP, computer vision, reinforcement, and deep learning. Matthew invites you to join him on his mission to simplify the future and to build AI for good.

저자(글) Rahul Kumar

Rahul Kumar is an AI scientist, deep learning practitioner, and independent researcher. His expertise in building multilingual NLU systems and large-scale AI infrastructures has brought him to Copenhagen, where he leads a large team of AI engineers as Chief AI Scientist at Jatana. Often invited to speak at AI conferences, he frequently travels between India, Europe, and the US where, among other research initiatives, he collaborates with The Intelligence Factory as NLP data science fellow. Keen to explore the ramifications of emerging technologies for his next book, he's currently involved in various research projects on Quantum Computing (QC), high-performance computing (HPC), and the brain-computer interaction (BCI).

저자(글) Abhishek Nagaraja

Abhishek Nagaraja was born and raised in India. Graduated Magna Cum Laude from the University of Illinois at Chicago, United States, with a Masters Degree in Mechanical Engineering with a concentration in Mechatronics and Data Science. Abhishek specializes in Keras and TensorFlow for building and evaluation of custom architectures in deep learning recommendation models. His deep learning skills and interest span computational linguistics and NLP to build chatbots to computer vision and reinforcement learning. He has been working as a Data Scientist for Skejul Inc. building an AI-powered activity forecast engine and engaged as a Deep Learning Data Scientist with The Intelligence Factory building solutions for enterprise clients.