Deep Learning with Python
by Ludovic SamperLanguages: English, French
Price (from): €2,000 / day
About the trainer:
Ludovic Samper is a data scientist who wants to show that there’s no magic in that field. He believes that to make good use of machine learning, engineers have to understand how and why data science works. He offers his expertise to companies who want to apply data science in their business. His training is based on programming so that participants can apply it easily within their business, and he also spends time on the underlying mathematics so that participants know how to tune it to their needs.
Ludovic Samper received his Ph.D. in computer science from Grenoble University in 2008.
Ludovic is a Python AI developer at Juripredis, a legal tech startup where he's building a jurisprudence search engine equipped with Artificial Intelligence (AI).
Previously, he was in Kuala Lumpur at the Center of Applied Data Science where he led the Data Science training team. He prepared and delivered training on every topic related to data science: programming, data cleaning, databases, probability theory, statistics, Machine Learning, Deep Learning. He taught all kinds of audience: from students to professionals and university lecturers and researchers in mathematics and computer science.
From 2008 to 2016, he worked as Research Engineer at Antidot.net, a search engine software editor. He was designing and implementing new features and solutions in the field of Natural Language Processing (NLP), Machine Learning (ML) and Information Retrieval (IR). For instance, he built a news portal, designed and implemented a model to improve the relevance of the search engine
Please login to send a message
Deep Learning with Python by Ludovic Samper
About the training
#hands on Training #Intuitive explanations of the founding principles #learn by doing #Artificial Intelligence #Natural Language Processing #Computer Vision #TensorFlow #Keras #Python #programming
Artificial Intelligence dates back from the 70s and includes multiple techniques, one of those which has proven to be efficient is Machine Learning (ML). In Machine Learning, a model is built based on many examples. With this model, it becomes possible to compute some predictions on new data. One of the machine learning techniques is deep learning whose foundations are neural networks. Deep learning models are today the state of the art solution to many real-world problems. This training aims at introducing deep learning for practitioners who want to benefit from those recent advances in Artificial Intelligence.
Python is one of the main general-purpose programming languages. Its popularity is due to the huge number of third-party libraries available. It has a large community of users among data scientists. This training will be based on TensorFlow (from Google), one of the most popular deep learning library and its high-level API (Keras).
This is practical training with many hands-on exercises. The training material is executable code which will run on each participant laptop.
The participant will get some intuitions and theoretical explanations about why things work. This way, she will be able to apply deep learning techniques to her problems and data soon after the training.
This training can be done in 3 days for people who already have some experience in Machine Learning. For people who discover the topic, 4 days is much better.
Learning outcomes
Implementation
The data practitioners (data scientists, developers, CTO) will learn how to implement deep learning models with TensorFlow.
Technical and scientific watch
The participants will get a very good sense of what can be achieved with deep learning and what still needs further research according to the scientific community.
Open-source
Python and the libraries used during the class are open sources. The code in the material can be reused soon after the class and doesn't need any proprietary tool to run.
Concepts
Deep learning concepts explained in simple English in this training don't depend on the libraries and programming languages. Attendees who use other technical stack in their company will still benefit from this session. All techniques presented here can be written in another deep learning framework.
Program
Machine Learning:
Curse of dimensionality
Overfitting
Cross validation
Foundations for Deep Learning:
Linear Regression
Gradient Descent algorithm
Ridge and Lasso Regularization
Logistic Regression
First neural network model: the multi layer perceptron
Simple deep learning models:
Introduction to TensorFlow and Keras
Loss functions and architectures for regression and classification.
Multi-class versus binary classification
Fighting overfitting: dropout, regularization.
Deep learning for Image Processing (computer vision):
Introduction to convolutions
First convolutional neural network
Data augmentation: how to use convolutional neural nets with small datasets
Transfer learning: using a pretrained model
Visualizing what convnet learn
Deep learning for Natural Language Processing:
Introduction to word embeddings
computing a problem specific word representation
using pre-trained word embeddings
recurrent neural networks: Simple recurrent networks, LSTM, GRU, bi-LSTM
Using recurrent networks for time-series
TensorFlow tool:
Tensorboard for model and training visualization
Main benefits
- #Wide collection of the biggest experts
- #Filters for all kinds of needs
- #User friendly platform
- #Fast and cheap
- #Highest level of proficiency
More trainings of the trainer