Name: ML conf EU 2020
Start: 2020-11-04T21:00:00.000Z
End: 2020-11-06T21:00:00.000Z

Varun

Hulya Karakaya

Kathinka

Robert

Ángel Fernández-Paniagua

Raquel Horta Bartomeu

Merla Nugas

Mayank Srivastava

Emre Hayırcı

Amar Shah

tal waserman

Luca Galli

Mehdi Mtir

Neha

Frans van Dunné

Liad Yosef

Håkan Silfvernagel

Descubre cómo aprovechar el aprendizaje automático en JavaScript utilizando TensorFlow.js en el navegador y más allá en esta charla rápida. Inspírate a través de un montón de prototipos creativos que empujan los límites de lo que es posible en el navegador web moderno (las cosas han avanzado mucho) y luego da tus primeros pasos con el aprendizaje automático en minutos. Al final de la charla, todos entenderán cómo reconocer un objeto de su elección que luego se puede utilizar de cualquier manera creativa que puedas imaginar. Se asume familiaridad con JavaScript, pero no se requiere experiencia en aprendizaje automático. ¡Ven y da tus primeros pasos con TensorFlow.js!

TensorFlow.js 101: Aprendizaje automático en el navegador y más allá

TensorFlow.js enables machine learning in the browser and beyond, with features like face mesh, body segmentation, and pose estimation. It offers JavaScript prototyping and transfer learning capabilities, as well as the ability to recognize custom objects using the Image Project feature. TensorFlow.js can be used with Cloud AutoML for training custom vision models and provides performance benefits in both JavaScript and Python development. It offers interactivity, reach, scale, and performance, and encourages community engagement and collaboration between the JavaScript and machine learning communities.

Discover how to embrace machine learning in JavaScript using TensorFlow.js in the browser and beyond in this speedy talk. Get inspired through a whole bunch of creative prototypes that push the boundaries of what is possible in the modern web browser (things have come a long way) and then take your own first steps with machine learning in minutes. By the end of the talk everyone will understand how to recognize an object of their choice which could then be used in any creative way you can imagine. Familiarity with JavaScript is assumed, but no background in machine learning is required. Come take your first steps with TensorFlow.js!

TensorFlow.js 101: ML in the Browser and Beyond

Jason Mayes

En esta charla comenzaré presentando los avances recientes en NLP que resultaron de la combinación de esquemas de aprendizaje por transferencia y arquitecturas de Transformer. La segunda parte de la charla estará dedicada a una introducción de las herramientas de código abierto lanzadas por HuggingFace, en particular nuestras bibliotecas Transformers, Tokenizers y Datasets y nuestros modelos.

Una introducción al aprendizaje por transferencia en NLP y HuggingFace

Transfer learning in NLP allows for better performance with minimal data. BERT is commonly used for sequential transfer learning. Models like BERT can be adapted for downstream tasks such as text classification. Handling different types of inputs in NLP involves concatenating or duplicating the model. Hugging Face aims to tackle challenges in NLP through knowledge sharing and open sourcing code and libraries.

In this talk I'll start introducing the recent breakthroughs in NLP that resulted from the combination of Transfer Learning schemes and Transformer architectures. The second part of the talk will be dedicated to an introduction of the open-source tools released HuggingFace, in particular our Transformers, Tokenizers and Datasets libraries and our models.

An Introduction to Transfer Learning in NLP and HuggingFace

Thomas Wolf

A menudo se piensa que para tener éxito con Machine Learning y Deep Learning, como una rampa de acceso a la Inteligencia Artificial, se necesita un profundo conocimiento en matemáticas y cálculo, así como algún tipo de doctorado. Pero no es así. Con APIs modernas como TensorFlow, gran parte de la complejidad se abstrae en bibliotecas preconstruidas, por lo que puedes centrarte en aprender. En esta sesión, Laurence Moroney, de Google, explicará cómo ha utilizado esto para crear cursos con cientos de miles de estudiantes, y a partir de ahí, cómo se creó un programa de certificación.

Enseñando ML y AI a los Programadores

The Talk discusses the current state of AI and the challenges faced in educating developers. Google's mission is to train 10 percent of the world's developers in machine learning and AI. They have developed specializations and training initiatives to make AI easy and accessible. The impact of AI education includes rigorous certification exams and partnerships with universities. The Talk also highlights the growth trends in the tech industry and the importance of AI skills. TensorFlow is recommended for its deployment capabilities, and practice is emphasized for building a career in machine learning.

Often it's thought that to be able to succeed with Machine Learning and Deep Learning, as an onramp to Artificial Intelligence, that you need a deep background in mathematics and calculus, as well as some form of PhD. But you don't. With modern APIs like TensorFlow, much of the complexity is abstracted away in pre-built libraries, so you can focus on learning. In this session, Laurence Moroney, from Google, will explain how he has used this to create courses with hundreds of thousands of students, and from there, how a certificate program was created.

Teaching ML and AI to Coders

Laurence Moroney

El dominio del Procesamiento del Lenguaje Natural ha visto una tremenda cantidad de investigación e innovación en los últimos años para abordar el problema de implementar soluciones de aprendizaje automático e IA de alta calidad utilizando texto natural. La clasificación de texto es una de las áreas más importantes en todos los sectores, como finanzas, medios de comunicación, desarrollo de productos, etc. Construir un sistema de clasificación de texto desde cero para cada caso de uso puede ser desafiante en términos de costos y recursos, considerando que se necesita una buena cantidad de datos para comenzar el entrenamiento. Aquí es donde entra en juego el concepto de aprendizaje transferido. Utilizar algunos de los modelos que han sido pre-entrenados en grandes cantidades de datos y ajustarlos según el problema en cuestión es la nueva forma de implementar soluciones de aprendizaje automático de manera eficiente sin pasar meses en la limpieza de datos. Esta charla destacará las formas de implementar BERT, recientemente lanzado, y ajustar el modelo base para construir un modelo de clasificación de texto eficiente. Se requiere un conocimiento básico de Python. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

El poder del aprendizaje transferido en NLP: Construye un modelo de clasificación de texto usando BERT

Transfer learning is a technique used when there is a scarcity of labeled data, where a pre-trained model is repurposed for a new task. BERT is a bidirectional model trained on plain text that considers the context of tokens during training. Understanding the baseline NLP modeling and addressing challenges like context-based words and spelling errors are crucial. BERT has applications in multiple problem-solving scenarios, but may not perform well in strict classification labels or conversational AI. Training BERT involves next sentence prediction and mass language modeling to handle contextual understanding and coherent mapping.

The domain of Natural Language Processing have seen a tremendous amount of research and innovation in the past couple of years to tackle the problem of implementing high quality machine learning and AI solutions using natural text. Text Classification is one such area that is extremely important in all sectors like finance, media, product development, etc. Building up a text classification system from scratch for every use case can be challenging in terms of cost as well as resources, considering there is a good amount of dataset to begin training with. Here comes the concept of transfer learning. Using some of the models that has been pre-trained on terates of data and fine-tuning it based on the problem at hand is the new way to efficiently implement machine learning solutions without spending months on data cleaning pipeline. This talk with highlight ways of implementing the newly launched BERT and fine tuning the base model to build an efficient text classifying model. Basic understanding of python is desirable. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Power of Transfer Learning in NLP: Build a Text Classification Model Using BERT

Jayeeta Putatunda

DeepPavlov Agent es un marco diseñado para facilitar el desarrollo de asistentes virtuales multi-habilidad escalables y listos para producción, sistemas de diálogo complejos y chatbots. Las características clave de DeepPavlov Agent incluyen (1) escalabilidad y confiabilidad en entornos de alta carga debido a la arquitectura de microservicios; (2) facilidad para agregar y orquestar habilidades conversacionales; (3) memoria compartida de estado de diálogo y anotaciones de NLP accesibles para todas las habilidades. DeepPavlov DREAM es una plataforma de socialbot con un diseño modular con componentes principales como anotadores, habilidades y selectores que se ejecutan como servicios independientes. Estos componentes se configuran e implementan utilizando contenedores Docker. Permite a los desarrolladores centrarse en el desarrollo de aplicaciones en lugar de enfocarse en los detalles intrínsecos de la configuración de infraestructura manual a nivel bajo. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

DeepPavlov Agent: Marco de código abierto para IA conversacional multihabilidad

The Pavlov Agent is an open source framework for multi-skill conversational AI, addressing the need for specific skills in different domains. The microservice architecture allows for scalability and skill reuse. The Deep Pavlov Library enables the creation of NLP pipelines for different skills. The Deep Pavlov Dream serves as a repository for skills and templates, while the Deployment Agent orchestrates all components for a seamless conversational experience. DeepLove.AI offers more flexibility and customization compared to Microsoft's LUIS service.

DeepPavlov Agent is a framework designed to facilitate the development of scalable and production-ready multi-skill virtual assistants, complex dialogue systems, and chatbots. Key features of DeepPavlov Agent include (1) scalability and reliability in the high load environment due to micro-service architecture; (2) ease of adding and orchestrating conversational skills; (3) shared dialogue state memory and NLP annotations accessible to all skills. DeepPavlov DREAM is a socialbot platform with a modular design with the main components such as annotators, skills and selectors run as independent services. These components are conﬁgured and deployed using Docker containers. It allows developers to focus on application development instead of focusing on the intrinsic details of the manual low-level infrastructure conﬁguration. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

DeepPavlov Agent: Open-source Framework for Multiskill Conversational AI

Mikhail Burtsev

Ven y descubre nuestro masterclass que te guiará a través de 3 recorridos comunes al usar TensorFlow.js. Comenzaremos demostrando cómo usar uno de nuestros modelos predefinidos, clases de JS muy fáciles de usar para trabajar rápidamente con ML. Luego veremos cómo volver a entrenar uno de estos modelos en minutos utilizando el aprendizaje por transferencia en el navegador a través de Teachable Machine y cómo se puede usar en tu propio sitio web personalizado. Finalmente, terminaremos con un hola mundo escribiendo tu propio código de modelo desde cero para hacer una regresión lineal simple y predecir los precios ficticios de las casas en función de sus metros cuadrados.<a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Prácticas con TensorFlow.js

JavaScript's versatility allows for machine learning across various environments, and TensorFlow.js enables machine learning in the browser. Pre-trained models and FaceMesh can be easily used in the web browser, ensuring privacy and compliance. Transfer learning and Cloud AutoML allow for retraining models and optimizing them for specific tasks. TensorFlow.js provides performance benefits and makes machine learning more accessible to a wider audience. The workshop covers hands-on demos, including training models with Teachable Machine and creating machine learning models from scratch using TensorFlow.js.

Come check out our workshop which will walk you through 3 common journeys when using TensorFlow.js. We will start with demonstrating how to use one of our pre-made models - super easy to use JS classes to get you working with ML fast. We will then look into how to retrain one of these models in minutes using in browser transfer learning via Teachable Machine and how that can be then used on your own custom website, and finally end with a hello world of writing your own model code from scratch to make a simple linear regression to predict fictional house prices based on their square footage. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Hands on with TensorFlow.js

La visualización de datos es una parte fundamental de la Ciencia de Datos. La charla comenzará con una demostración práctica (utilizando pandas, scikit-learn y matplotlib) de cómo confiar solo en estadísticas resumidas y predicciones puede dejarte ciego ante la verdadera naturaleza de tus conjuntos de datos. Haré hincapié en que las visualizaciones son cruciales en cada paso del proceso de Ciencia de Datos y, por lo tanto, los Jupyter Notebooks definitivamente pertenecen a la Ciencia de Datos. Luego veremos cómo la mantenibilidad es un desafío real para los Jupyter Notebooks, especialmente cuando se intenta mantenerlos bajo control de versiones con git. Aunque existen numerosas herramientas de calidad de código para scripts de Python (flake8, black, mypy, etc.), la mayoría de ellas no funcionan en Jupyter Notebooks. Con este fin, presentaré nbQA, que permite ejecutar cualquier herramienta estándar de calidad de código de Python en un Jupyter Notebook. Por último, demostraré cómo usarlo dentro de un flujo de trabajo que permite a los profesionales mantener la interactividad de sus Jupyter Notebooks sin tener que sacrificar su mantenibilidad. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

¡Nunca vuelvas a tener un Jupyter Notebook inmantenible!

Jupyter Notebooks are important for data science, but maintaining them can be challenging. Visualizing data sets and using code quality tools like NBQA can help address these challenges. Tools like nbdime and Precommit can assist with version control and future code quality. Configuring NBQA and other code quality tools can be done in the PyProject.toml file. NBQA has been integrated into various projects' continuous integration workflows. Moving code from notebooks to Python packages should be considered based on the need for reproducibility and self-contained solutions.

Data visualisation is a fundamental part of Data Science. The talk will start with a practical demonstration (using pandas, scikit-learn, and matplotlib) of how relying on summary statistics and predictions alone can leave you blind to the true nature of your datasets. I will make the point that visualisations are crucial in every step of the Data Science process and therefore that Jupyter Notebooks definitely do belong in Data Science. We will then look at how maintainability is a real challenge for Jupyter Notebooks, especially when trying to keep them under version control with git. Although there exists a plethora of code quality tools for Python scripts (flake8, black, mypy, etc.), most of them don't work on Jupyter Notebooks. To this end I will present nbQA, which allows any standard Python code quality tool to be run on a Jupyter Notebook. Finally, I will demonstrate how to use it within a workflow which lets practitioners keep the interactivity of their Jupyter Notebooks without having to sacrifice their maintainability. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Never Have an Unmaintainable Jupyter Notebook Again!

Marco Gorelli

La fabricación y seguimiento de chips y equipos es una tarea difícil dada la estricta adherencia a los estándares de calidad y procesos como los controles de seis sigma. En esta sesión, analizaremos problemas clave del mundo real en la industria de semiconductores y fabricación, y posibles metodologías en las que aprovechamos una combinación de técnicas tradicionales de visión por computadora y las combinamos con el poder del aprendizaje profundo de transferencia y el aprendizaje profundo. Cubriremos dos casos de uso principales de la industria: Detección automática de defectos a nanoescalaAgrupación de defectos a nanoescala <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Aprendizaje profundo de transferencia para la visión por computadora

Today's Talk focuses on deep transfer learning for Computer Vision in the semiconductor manufacturing industry, specifically defect classification. The speakers discuss using a hybrid classification system with pre-trained models and image augmentation for accurate defect detection. They also explore the use of unsupervised learning, leveraging clustering algorithms and pre-trained models like ResNet-50, for defect analysis without prior knowledge. The process is reproducible, user-friendly, and provides accurate cluster results, with potential for future supervised learning applications.

Chip and equipment manufacturing and tracking is a tough task given the strict adherence to quality standards and processes like six-sigma control checks. In this session, we will be looking at key real-world problems from Semiconductor &amp; Manufacturing, and possible methodologies where we leveraged a combination of traditional computer vision techniques and coupled it with the power of deep transfer learning and machine \ deep learning. We will be covering two main use-cases from the Industry: Automatic Defect detection at NanoscaleDefect Clustering at Nanoscale <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Deep Transfer Learning for Computer Vision

Dipanjan Sarkar

Sachin Dangayach

En muchas aplicaciones del mundo real, la calidad y curación de los datos y el conocimiento del dominio juegan un papel mucho más importante en la construcción de modelos exitosos que la creación de técnicas de procesamiento complejas y la optimización de hiperparámetros. Por lo tanto, una caja de herramientas de aprendizaje automático debe permitir a los usuarios comprender tanto los datos como el modelo, y no cargar al profesional con la selección de pasos de preprocesamiento y hiperparámetros. La biblioteca dabl es un primer paso en esta dirección. Proporciona rutinas de visualización automáticas y capacidades de inspección de modelos al tiempo que automatiza la selección de modelos. dabl contiene tipos de gráficos que no están disponibles en las bibliotecas estándar de Python hasta ahora, así como algoritmos novedosos para seleccionar visualizaciones interesantes. Se utilizan heurísticas para seleccionar el preprocesamiento adecuado para el aprendizaje automático, mientras que se utilizan algoritmos de selección de cartera de vanguardia para una búsqueda eficiente de modelos e hiperparámetros. dabl también proporciona un fácil acceso a herramientas de evaluación e inspección de modelos proporcionadas por scikit-learn. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Dabl: Aprendizaje Automático Automático con la Participación Humana

This talk introduces Dabble, a library that allows data scientists to iterate quickly and incorporate human input into the machine learning process. Dabble provides tools for each step of the machine learning workflow, including problem statement, data cleaning, visualization, model building, and model interpretation. It uses mosaic plots and pair plots to analyze categorical and continuous features. Dabble also implements a portfolio-based automatic machine learning approach using successive halving to find the best model. The future goals of Dabble include supporting more feature types, improving the portfolio, and building explainable models.

In many real-world applications, data quality and curation and domain knowledge play a much larger role in building successful models than coming up with complex processing techniques and tweaking hyper-parameters. Therefore, a machine learning toolbox should enable users to understand both data and model, and not burden the practitioner with picking preprocessing steps and hyperparameters. The dabl library is a first step in this direction. It provides automatic visualization routines and model inspection capabilities while automating away model selection. dabl contains plot types not available in standard python libraries so far, as well as novel algorithms for picking interesting visualizations. Heuristics are used to select appropriate preprocessing for machine learning, while state-of-the-art portfolio selection algorithms are used for efficient model and hyperparameter search. dabl also provides easy access to model evaluation and model inspection tools provided scikit-learn. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Dabl: Automatic Machine Learning with a Human in the Loop

Andreas Müller

Esta charla mostrará cómo se ha desarrollado un modelo de detección de fraudes basado en los datos de las sesiones de navegación de los diferentes usuarios. Se han utilizado herramientas como PySpark y Spark ML en esta iniciativa debido a la gran cantidad de datos. El modelo creado fue capaz de identificar un grupo de características que cubrían el 10% de las sesiones totales, de las cuales el 88% se consideraron fraudulentas. Esto permite a los analistas dedicar más tiempo a los casos de mayor riesgo. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Análisis de Sesiones del Navegador: La Clave para la Detección de Fraudes

Blue Tab Solutions specializes in advanced analytics and big data, and recently improved financial fraud detection using Spark and the CRISPM methodology. They discovered insights like the correlation between fraudulent sessions and the mobile cast page accessed from the web application. The models created using decision trees, random forest classifiers, and gradient boosting classifiers were validated using the area under the ROC curve. The GVT classifier yielded the best result with a score of 0.94. Regular training is necessary for accurate models, and the next steps involve real-time action when fraud is detected.

This talk will show how a fraud detection model has been developed based on the data from the browsing sessions of the different users. Tools such as PySpark and Spark ML have been used in this initiative due to a large amount of data. The model created was able to identify a grouping of characteristics that covered 10% of the total sessions in which 88% were deemed fraudulent. This allows analysts to spend more of their time on higher-risk cases. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Browser Session Analytics: The Key to Fraud Detection

Javier Alcaide Pérez

Después de esta charla, sabrás cómo escribir la canción perfecta para tu cantante favorito/a. Esto no es un retiro de composición de canciones, sino una charla sobre algunas de las características líricas, estructurales, armónicas y melódicas que incluye el análisis de canciones. Discutiremos la extracción de estructuras de canciones utilizando herramientas de procesamiento de lenguaje natural y análisis de repetición, características musicales y cómo utilizar todas esas características para predecir qué canciones se adaptan mejor a qué artista. Asiste a esta charla para descubrir cuál es el futuro que la industria musical puede lograr con el aprendizaje automático.<a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

¿Puedes Cantar con Todas las Voces de las Características?

This Talk discusses the role of repetition in songwriting and how it has become more prevalent over the years. The use of string metrics, such as the Levenstein distance, allows for the analysis of similarity between segments of songs. A similarity threshold of 70% is used to determine if segments are considered similar. Overall, the Talk explores the importance of repetition in creating successful songs and the use of analytical tools to measure similarity.

After this talk, you will know how to write the perfect song for your favourite singer! This is not a songwriting retreat but a talk about some of the lyrical, structural, harmonic, and melodic features song analysis includes. We will discuss the extraction of song structures using NLP tools and repetition analysis, musical features, and how to use all of those features to predict which songs fit which artist the best. Attend this talk to discover what is the future that the music industry can achieve with machine learning. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Can You Sing with All the Voices of the Features?

Yama Anin Aminof

La adopción de IA se ha visto frenada por los desafíos involucrados en obtener modelos de rendimiento, que requieren experiencia y esfuerzo significativos, y el número limitado de profesionales con experiencia en aprendizaje automático. El aprendizaje automático automatizado (AutoML) elimina los pasos rutinarios en el flujo de trabajo de aprendizaje automático, lo que permite a los expertos en el dominio sin experiencia en aprendizaje automático construir modelos iniciales sólidos y permite a los profesionales experimentados enfocarse en la optimización manual adicional del modelo. Esta charla describe el nivel de automatización disponible para los diversos pasos y demuestra AutoML con un clasificador de actividades humanas basado en datos de sensores de acelerómetro. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Ampliando la adopción de IA con AutoML

AutomL simplifies the complexity of building machine learning models, allowing engineers to focus on the hard problems and applications. It enables the solving of problems that wouldn't be feasible otherwise. The three-step AutomL approach by MathWorks includes wavelet scattering for feature extraction. AutoML also enables feature selection and model optimization for memory and power-limited embedded systems. MATLAB can translate to low-level code for deployment.

Adoption of AI has been slowed the challenges involved in obtaining performant models, which require significant expertise and effort, and the limited number of practitioners with machine learning expertise. Automated machine learning (AutoML) eliminates the routine steps in the machine learning workflow, thus empowering domain experts without machine learning background to build good initial models, and allowing experienced practitioners to focus additional manual model optimization. This talk describes the extent of automation available for the various steps and demonstrates AutoML with a classifier for human activities based on accelerometer sensor data. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Broadening AI Adoption with AutoML

Bernhard Suhm

La mayoría de las personas estarían de acuerdo en que R es un lenguaje popular para el análisis de datos. Quizás menos conocido es que R tiene un buen soporte para la ejecución paralela en una sola CPU a través de paquetes como future. En esta presentación hablaremos sobre nuestra experiencia escalando los procesos de R aún más, ejecutando R en paralelo en contenedores docker utilizando Kubernetes. Los robots generan grandes cantidades de datos de sensores y otros; extraer la información y conocimientos adecuados de esto requiere un procesamiento significativamente mayor que el que se puede abordar en un entorno de ejecución único. Ante la tarea de preprocesar varios cientos de GB de datos de archivos json comprimidos, utilizamos Pachyderm para escribir canalizaciones de datos para ejecutar la preparación de datos en paralelo, utilizando contenedores multinúcleo en un clúster de Kubernetes. Al final de la charla habremos desmentido el mito de que R no se puede utilizar en producción a gran escala. Incluso si no utiliza R, habrá visto un caso de uso para escalar el análisis independientemente de su lenguaje de elección. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Procesamiento de Datos de Robots a Escala con R y Kubernetes

The Talk discusses the challenges of managing and analyzing the increasing volume of data gathered from robots. It highlights the importance of data extraction and feature engineering in analyzing what happens before a failure. The use of Kubernetes and Packyderm for data management and automatic updates in the pipeline is mentioned. The parallelization of R scripts and the scalability of large clusters for data collection and processing are emphasized. The Talk also mentions the use of AI at the robot fleet level for unlocking new opportunities.

Most people would agree that R is a popular language for data analysis. Perhaps less well known is that R has good support for parallel execution on a single CPU through packages like future. In this presentation we will talk about our experience scaling up R processes even further running R in parallel in docker containers using Kubernetes. Robots generate massive amounts of sensor and other data; extracting the right information and insights from this requires significant more processing than can be tackled on a single execution environment. Faced with a preprocessing job of several hundred GB of data of compressed json line files, we used Pachyderm to write data pipelines to run the data prep in parallel, using multicore containers on a kubernetes cluster. By the end of the talk we will have dispelled the myth that R cannot be used in production at scale. Even if you do not use R, you will have seen a use case to scale up analysis regardless of your language of choice. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Processing Robot Data at Scale with R and Kubernetes

Como científico de IA y desarrollador, he estado involucrado en aplicaciones de IA durante muchos años, especialmente enfocándome en la detección y reconocimiento de objetos. Me encanta pensar que podemos ser creativos al diseñar redes neuronales. Podemos entrenarlas de manera supervisada, no supervisada, semi o auto-supervisada, y esto nos da la posibilidad de imitar el cerebro humano en un dominio específico. Sin embargo, en las aplicaciones de visión, todavía hay cosas en las que la IA carece y seguirá careciendo sin el conocimiento de visión por computadora. La visión por computadora ha estado resolviendo problemas de detección y reconocimiento durante muchos años. Sin embargo, en la última década, parece que la IA se ve como un reemplazo de la visión por computadora. La IA puede encontrar el modelo óptimo para un tipo específico de conjunto de datos y puede lograr una mejor generalización. La IA puede ser diseñada de tal manera que pueda aprender de por vida, lo que también brinda la posibilidad de crear modelos que funcionen mejor a medida que se usan durante más tiempo. Sin embargo, un sistema de visión de IA carecerá de capacidades sin el conocimiento de visión por computadora. En primer lugar, requerirá un conjunto de datos muy grande para entrenar el modelo, lo que puede ser costoso o incluso imposible. Por otro lado, los sistemas de visión por computadora pueden ser modelados solo utilizando una imagen de plantilla dibujada a mano. El entrenamiento de modelos de IA también requiere GPUs. Sin embargo, no quiero animar a todos a entrenar modelos de IA para resolver cualquier problema simple que podría resolverse fácilmente con visión por computadora. Por último, pero no menos importante, conocer la visión por computadora, el aprendizaje automático y especialmente los métodos de ingeniería de características ayuda a diseñar modelos híbridos que pueden ser más robustos ante ataques adversarios o condiciones cambiantes. En esta charla, presentaré brevemente cómo se puede utilizar la visión por computadora (especialmente utilizando la biblioteca OpenCV) y el aprendizaje automático para crear modelos de detección y reconocimiento. Sería útil tener algo de experiencia con Python, Jupyter Notebook y algo de conocimiento en aprendizaje automático para obtener más beneficios de esta charla. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Visión por Computadora utilizando OpenCV

Today's Talk explores image processing, computer vision, and their combination with machine learning. Image processing involves manipulating images, while computer vision extracts valuable information from images. Histograms are crucial in image processing as they represent the distribution of brightness values. Various image processing techniques can be used, such as thresholding and convolution. Computer vision techniques focus on extracting important features for object recognition and can be hand-tailored. Audio processing is not the focus of OpenCV, but TensorFlow libraries may be more suitable. Understanding the algorithms behind the code is important for robustness and effective debugging. Computer vision has applications in healthcare for cancer recognition and in agriculture for plant health monitoring.

As an AI scientist and a developer, I have been engaged with AI-applications for many years especially focusing on object detection and recognition purposes. I love thinking that we can get creative in designing neural networks. We can train them supervised, unsupervised, semi or self-supervised, and this gives possibilities to mimic the human brain in a narrow domain. However, in vision applications, there are still things where AI is lacking and will be lacking without computer vision knowledge. Computer vision has been solving detection and recognition problems for many years. However, in the last decade, it seems like AI is seen as a replacement of computer vision. AI can find the optimal model for a specific type of data set and it might achieve generalization better. AI can be designed in a way that it can learn life-long which also brings possibilities of creating models which serve better when they are used longer. However, an AI vision system will be lacking capabilities without computer vision knowledge. First of all, it will require a very big data set to train the model what can be expensive or even not possible. On the other hand, computer vision systems can be modeled only using a hand-drawn template image. Training AI models also requires GPUs. Nevertheless, I do not want to encourage everyone to train AI models for solving any simple problem which could be solved easily computer vision. Last but not least, knowing computer vision, machine learning and especially feature engineering methods helps to design hybrid models that might be more robust to adversarial attacks or changing conditions. In this lecture, I will briefly introduce how computer vision (especially using the OpenCV library) and machine learning can be used for creating detection and recognition models. Some experience with python, jupyter notebook and some machine learning background would be useful to get more benefits from this lecture. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Computer Vision Using OpenCV

Beril Sirmacek

¿Eres un Ingeniero de Software al que se le asignó la tarea de implementar un modelo de aprendizaje automático o aprendizaje profundo por primera vez en tu vida? ¿Te preguntas qué pasos seguir y cómo se diferencia el software impulsado por IA del software tradicional? Entonces este es el masterclass adecuado al que asistir. Internet ofrece miles de artículos y cursos gratuitos que muestran lo fácil que es entrenar e implementar un modelo de IA simple. Al mismo tiempo, en la realidad es difícil integrar un modelo real en la infraestructura actual, depurarlo, probarlo, implementarlo y monitorearlo correctamente. En este masterclass, te guiaré a través de este proceso compartiendo consejos, trucos y herramientas de código abierto favoritas que te facilitarán mucho la vida. Así que al final del masterclass, sabrás por dónde empezar tu viaje de implementación, qué herramientas utilizar y qué preguntas hacer. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

La Guía del Autoestopista de la Galaxia de Ingeniería de Aprendizaje Automático

This Workshop covers the difference between AI-powered software and traditional software, the theory of AI-powered software, MLOps, AI pipelines, and the ML platform. It also discusses the challenges of AI versioning and MLOps, the importance of model training and versioning, and the process of model evaluation and deployment. The Workshop explores pipeline automation, ML metadata store, AI platform and deployments, and options for offline serving in batch mode. It also covers working with batch models, defining input and output schemes, and performing batch inference and deployment. The Workshop concludes with insights on offline and online deployment, A-B testing for online deployments, and tools for data versioning and ML deployment.

Are you a Software Engineer who got tasked to deploy a machine learning or deep learning model for the first time in your life? Are you wondering what steps to take and how AI-powered software is different from traditional software? Then it is the right workshop to attend. The internet offers thousands of articles and free of charge courses, showing how it is easy to train and deploy a simple AI model. At the same time in reality it is difficult to integrate a real model into the current infrastructure, debug, test, deploy, and monitor it properly. In this workshop, I will guide you through this process sharing tips, tricks, and favorite open source tools that will make your life much easier. So, at the end of the workshop, you will know where to start your deployment journey, what tools to use, and what questions to ask. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

The Hitchhiker's Guide to the Machine Learning Engineering Galaxy

Alyona Galyeva

¿Qué pasaría si pudieras realizar aprendizaje automático en el borde, es decir, en tu dispositivo móvil? Esto significaría que ya no necesitarías el viaje de ida y vuelta al servidor, ningún dato saldrá del dispositivo y ni siquiera necesitarías una conexión a internet. En esta sesión obtendrás una introducción a TensorFlow Lite para que puedas usarlo en tus propios proyectos. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Aprendizaje automático en el borde utilizando TensorFlow Lite

Håkan Silvernagel introduces TensorFlow Lite, an open-source deep learning framework for deploying machine learning models on mobile and IoT devices. He highlights the benefits of using TensorFlow Lite, such as reduced latency, increased privacy, and improved connectivity. The Talk includes a demonstration of object recognition capabilities and a real-world example of using TensorFlow Lite to detect a disease affecting farmers in Tanzania. References to official TensorFlow documentation, Google IO conference, and TensorFlow courses on Coursera are provided.

What if you could perform machine learning on the edge, i.e on your mobile device? This would mean that you no longer would need the roundtrip to the server, no data will leave the device and you don't even need an internet connection. In this session you will get an introduction to TensorFlow Lite so that you can use it in your own projects. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Machine Learning on the Edge Using TensorFlow Lite

Este masterclass será tanto una introducción suave al Aprendizaje Automático, como un ejercicio práctico de uso de la nube para entrenar modelos de aprendizaje automático simples y no tan simples. Comenzaremos utilizando ML Automático para entrenar el modelo para predecir la supervivencia en el Titanic, y luego pasaremos a tareas de aprendizaje automático más complejas como la optimización de hiperparámetros y la programación de series de experimentos en el clúster de cómputo. Finalmente, mostraré cómo Azure Machine Learning se puede utilizar para generar pinturas artificiales utilizando Redes Generativas Adversarias, y cómo entrenar un modelo de preguntas y respuestas de lenguaje en documentos de COVID para responder preguntas relacionadas con COVID.<a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Introducción al Aprendizaje Automático en la Nube

Azure Machine Learning is a powerful tool for both experienced data scientists and beginners, providing tools for training machine learning models and scheduling experiments. It offers flexibility in separating compute from storage and includes features like AutoML and the ability to create pipelines using the Designer tool. The Workshop also covers running experiments remotely on clusters, hyperparameter optimization, training generative adversarial networks, and using DeepPavlov for question-answering models. Azure ML can be used for low code and no code scenarios, making it a versatile platform for machine learning tasks.

This workshop will be both a gentle introduction to Machine Learning, and a practical exercise of using the cloud to train simple and not-so-simple machine learning models. We will start with using Automatic ML to train the model to predict survival on Titanic, and then move to more complex machine learning tasks such as hyperparameter optimization and scheduling series of experiments on the compute cluster. Finally, I will show how Azure Machine Learning can be used to generate artificial paintings using Generative Adversarial Networks, and how to train language question-answering model on COVID papers to answer COVID-related questions. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Introduction to Machine Learning on the Cloud

Dmitry Soshnikov

TensorFlow ha construido una sólida base para diversas aplicaciones de aprendizaje automático, sobre la cual el ecosistema de Keras puede realmente aumentar la productividad de los desarrolladores en la construcción de soluciones de aprendizaje automático. Keras tiene una API simple y arbitrariamente flexible para construir y entrenar modelos. Sin embargo, todavía necesitamos mucho trabajo manual para ajustar los hiperparámetros. Afortunadamente, con Keras Tuner, podemos automatizar el proceso de ajuste de hiperparámetros con modificaciones menores en el código para construir y entrenar los modelos. Para aumentar aún más la productividad, presentamos AutoKeras, que automatiza completamente el proceso de construcción de modelos, entrenamiento y ajuste de hiperparámetros. Reduce drásticamente la cantidad de conocimientos previos necesarios para utilizar el aprendizaje automático en algunas tareas comunes. Todo lo que necesitas es definir la tarea y proporcionar los datos de entrenamiento. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Aumenta la productividad con el ecosistema de Keras

This Talk introduces the TensorFlow in Keras ecosystem and highlights its features, including tensor manipulations, automatic differentiation, and deployment. It also discusses the workflow and automation of hyperparameter tuning with Keras Tuner and AutoKeras. The Talk emphasizes the simplicity and productivity of using AutoKeras, which supports various tasks and advanced scenarios. It also mentions the challenges beginners face and provides resources for learning. Lastly, it touches on the use of TensorFlow and Keras in the research domain and the customization options in AutoKeras, including time series forecasting.

TensorFlow has built a solid foundation for various machine learning applications, on top of which the Keras ecosystem can really boost the productivity of the developers in building machine learning solutions. Keras has a simple and arbitrarily flexible API for building and training models. However, we still need a lot of manual work to tune the hyperparameters. Fortunately, with Keras Tuner, we can automate the hyperparameter tuning process with minor modifications to the code for building and training the models. To further boost the productivity, we introduce AutoKeras, which fully automates the model building, training, and hyperparameter tuning process. It dramatically reduces the amount of prior knowledge needed of using machine learning for some common tasks. All you need is to define the task and to provide the training data. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Boost Productivity with Keras Ecosystem

Haifeng Jin

Las matemáticas elegantes y graciosas hacen una portada de libro de texto genial, pero el interior de esos mismos libros suele ser una ingeniería fría y seca. Es importante mezclar la teoría de la innovación con la emoción de la practicidad, y a través de la composición de estos elementos encontramos la innovación. En esta charla, te mostraré desde una perspectiva de ingeniería cómo explorar, equilibrar y, en última instancia, embotellar el éxito mecanizado. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

La Revolución de la Evolución

The Talk discusses the challenges of implementing software solutions and the need for abstractions. It emphasizes the importance of innovation and implementing once to avoid complexity. The use of Brain.js in machine learning research and its practical applications are highlighted. The talk also mentions the benefits of using JavaScript and GPU.js for graphics processing. Overall, the Talk encourages simplicity, efficiency, and collaboration in software development.

Elegant and graceful mathematics make a cool textbook cover, but the inside of those same books are usually dry cold engineering. It's important to mix the theory of innovation with the excitement of practicality, and through the composition of these elements we find innovation. In this talk, I'll show you from an engineering perspective how to explore, balance, and ultimately bottle machined success. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

The Evolution Revolution

Robert Plummer

Esta charla será una guía paso a paso sobre cómo utilizar el aprendizaje automático para reemplazar un sistema basado en reglas para los consumidores. Discutiremos cuándo está bien usar ML, cómo construir estos modelos con datos inteligentes, evaluarlos sin conexión y finalmente cómo validar esta evaluación para implementar estos modelos en sistemas de producción. Además, ilustraremos diversas estrategias de autoaprendizaje/aprendizaje interactivo que se pueden utilizar en sistemas de producción para automatizar cómo los modelos se enseñan a sí mismos para mejorar.<a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

Cómo convertir cualquier producto en un producto de aprendizaje automático

In this Talk, an ML engineer from Facebook shares insights on when to use ML and a successful use case from Facebook. The speaker discusses the process of using ML for Facebook Portal calls, including data collection and model selection. The importance of precision and recall in ML models is emphasized, as well as the need for online evaluation and active learning. The Talk also touches on the challenges of data protection and label delay in ML model development.

This talk will be a walkthrough of utilizing machine learning to replace a rule based system for consumers. We will discuss when is it okay to use ML, how to build these models with intelligent data, evaluate these offline and finally how to validate this evaluation to land these models in production systems. Furthermore, we will illustrate various self-learning/interactive-learning strategies that can be used for production systems to automate how models teach themselves to become better. <a href="https://www.froala.com/wysiwyg-editor?pb=1"> </a>

ML conf EU 2020

The Machine Learning conference for software developers