Trending this week: Build vison-language models using X-modaler; Use example weighting strategy for deep learning; Get notified the second your training is done using MLNotify.
Every week we analyze the most discussed topics on Twitter by Data Science & AI influencers.
The following topics, URLs, resources, and tweets have been automatically extracted using a topic modeling technique based on Sentence BERT, which we have enhanced to fit our use case.
This week, Data Science and AI influencers on Twitter have talked about:
- Amazing DL resources
- ML Updates
- Potential of AI
Amazing DL Resources
This week, data science and AI influencers have shared amazing resources on deep learning and deep reinforcement learning.
Dr. Georgina Cosma has shared a link to the Github project X-modaler, a versatile and high-performance codebase for cross-modal analytics between vision and language in multimedia field. This codebase is an implementation of the original paper “X-modaler: A Versatile and High-performance Codebase for Cross-modal Analytics”, and unifies comprehensive high-quality modules in state-of-the-art vision-language techniques, which are organized in a standardized and user-friendly fashion. X-modaler enables flexible implementation of state-of-the-art algorithms for image captioning, video captioning, and vision-language pre-training, aiming to facilitate the rapid development of research community. Also, it can be simply extended to power startup prototypes for other tasks in cross-modal analytics, including visual question answering, visual commonsense reasoning, and cross-modal retrieval.
ipfconline has shared:
A link to an amazing online Free Course in Deep Reinforcement Learning. This course is a series of articles and videos where you’ll master the skills and architectures you need, to become a deep reinforcement learning expert. It guides you step by step through the different chapters in implementing awesome DL agents with Tensorflow and PyTorch. All the codes are available in the associated Github project.
A student doctoral thesis discussing the strategy of Example weighting for deep representation learning. Example weighting, which becomes an active research field, helps to learn more robust and discriminative representations using deep supervised learning, as it provides better optimization and regularisation. this paper focuses on two learning tasks: learning to rank, and learning to classify. This study claims that example weighting is an effective approach for addressing the challenge of differentiating trusted and error patterns in the training of a deep learning model, and avoid fitting the error transformation. Indeed, when a training dataset is clean, naively assigning higher weights to harder examples works well. However, when the dataset contains both meaningful and wrong information, a model learns meaningful patterns before fitting random errors. Which ultimately misleads the model.
KDnuggets have shared:
An article on an Overview of Different Approaches to Deploying Machine Learning Models in Production. This post introduces the different methods for putting machine learning models into production in order to help you determine which method is best for which use case. This article details ML industrialization approaches like one-off training, batch training, and real-time training. Also, it explains how these approaches can be used to solve different use cases, and it provides the most popular technologies associated with them. Finally, this post emphasizes that understanding specific use cases, the team’s technical and analytics maturity, the overall organization structure, and its’ interactions, help come to the right approach for deploying predictive models to production.
The application MLNotify, an open-source tool that watches model training for you and sends a notification once training is complete. MLNotify is a Python library that notifies you the second your training is done via the web, mobile, and email notifications — with just one import line. To make it possible, MLNotify hooks into the fit() method of popular ML libraries, and notifies upon completion of the method execution. The associated code is available on GitHub.Using MLNotify,You can enable web, mobile, or email notifications to know exactly when your training finishes.
Mike Tamir has shared an article that provides an organization of various kinds of biases that can occur in the ML pipeline starting from dataset creation and problem formulation to data analysis and evaluation. It introduces 10 types of bias like sampling, measurement, label, human evaluation, sample selection biases, and others. It highlights the challenges associated with the design of bias-mitigation strategies, and it outlines some best practices suggested by researchers. Finally, a set of guidelines is presented that could aid ML developers in identifying potential sources of bias, as well as avoiding the introduction of unwanted biases.
Potential of AI
This week many data science and artificial influencers talked about the potential of artificial intelligence across a wide plethora of domains.
Terence Mills shared an article on artificial intelligence and what it means for education. Andreas Staub has also shared an article on the significant roles of artificial intelligence in the education sector. The article talks about how proper implementation of AI in the education sector can lead to lessening workload for educators and learners, creating global learning opportunities for the students, and personalizing training for students.
Andreas Staub also shared a research paper on the use and impact of artificial intelligence on climate change adaptation in Africa. The paper talks as the world enter the fourth industrial revolution, the adoption of advanced technologies such as artificial intelligence (AI) introduces complex challenges and opportunities for the now-inevitable and as-yet-undetermined issues of climate change.
Dr. Mark van Rijmenam shared a video in which he talks about combining AI to analyze medical results and nano-bots to detect and treat mutations inside a cell, without instruction, we suddenly see a near-future where digital healthcare speeds up exponentially.
Finally, Terence Millsshared an article on how modern AI-enabled automation unlocks full human potential. It mentions three main aspects of AI-enabled automation — Understand the difference between task-based, role-based, and function-based automation. Pinpoint the best automation opportunities. Evaluate and select the best automation solution and approach.