7 of The Coolest Machine Learning Topics of 2021
Machine LearningModelingWest 2021posted by Sheamus McGovern November 1, 2021 Sheamus McGovern
ODSC West 2021 is only a few short weeks away and the agenda is live so now is a good time for a quick pass on what’s cool and trending at this upcoming event. Here are some of our top picks to keep machine learning and data science practitioners abreast of trending topics in the field with these popular machine learning topics.
Machine Learning Safety
Pause for a moment to realize the number of machine learning models trained on crowdsourced data from social media and the web, and realize how easy it is to poison training data. In fact, this Microsoft paper from last year identifies it as a top concern (p.2). Driven by foundational models, large-scale models, and autonomous systems, ML safety is quickly becoming a broad topic encompassing many areas of AI and ML. Adversarial attacks, backdoor model vulnerabilities, real-world deployment tail risks, risk monitoring, and boosting defense are a few of the topics to fall under the ML safety umbrella. Expect to hear a lot more on this fast-trending topic.
Massive trained models such as GPT-3 and BERT have been all the rage over the last few years, deserving acclaim for their breakthrough accomplishments. Termed foundational models by the Stanford HAI center, these models have come under new scrutiny. A single model can be employed across many applications, amplifying the challenges and risks of machine learning system design. Understanding the power, opportunity, and risk associated with these models will be fundamental to building responsible AI. ODSC West will feature many of these models, dissecting their capabilities and vulnerabilities.
Machine Learning Observability
MLOPs, AIOPs, DataOps. Any acronym can be the flavor of the moment thanks to heavy industry investment and a wall of VC funding. Dig a little deeper and you’ll notice a lot of unsolved problems in what, acronyms aside, is the ML systems engineering space. Once deployed to production, ML engineers need to monitor for model drift, data drift, data degradation, model improvement, and of course error detection. Observability is not just for real-time systems or even production environments. Applying the discipline of ML observability can identify problems early and display the belief that some ML lifecycles are static. ODSC West will have one of its strongest lineups of data engineering and MLOps sessions to date.
Deep Generative Learning
Deep Generative models (DGMs) have been around for a while now and received a lot of attention for generating deep fakes, but they have also been successfully used in hidden Markov models, GANs, bayesian networks, autoregressive models, and more. DGMs are neural nets with many hidden layers trained to high-dimensional probability distributions using a large number of samples. Despite these early successes, the broader use of DGMs is still in the early stages. It’s one of the hottest research topics in many of the top universities as researchers seek better ways to design and train these models. With this continued focus from some of the industry’s top minds, we can look forward to more breakthroughs and wider adoption for practical applications.
Privacy-Preserving Machine Learning and Differential Privacy
Permitting multiple organizations to collaboratively build, train, and deploy machine learning models without jeopardizing data privacy continues to gain importance. Responsible AI is a broad term employed by industry while practitioners prefer to focus on the many challenging issues of ensuring true end-to-end, privacy-preserving machine learning models. The focus is now on all states of the machine learning life cycle, including understanding privacy as it relates to training data, model inputs, model weights, model outputs, and model monitoring. Additionally, the field is evolving beyond basic differential privacy techniques, such as purposely introducing statistical or other types of noise to model inputs and outputs. Machine learning practitioners will find the latest on this topic at ODSC West.
Deep Learning-Based Natural Language Processing
NLP continues to enjoy a resurgence of interest in the industry thanks to developments in the last few years, including transfer learning and transformer models. New techniques combining supervised learning and unsupervised learning are gaining traction and advances continue to be made employing various deep learning techniques. Recursive Neural Networks and Recurrent Neural Networks’ (RNNs) specialty for processing sequential information, such as text, make them especially useful for NLP models. Deep Generative Models (DGMs), as previously mentioned, have led to massive breakthroughs in NLP. ODSC West will have many exciting sessions on NLP.
Machine Learning for Cybersecurity
Given the increasing importance of machine learning safety, it is essential that engineers and experts in AI broaden their knowledge of cybersecurity. In addition, cybersecurity is a field that is rapidly growing thanks to the deployment of machine learning tools and methods. Experts are employing machine learning to help predict and craft better threat incident response, monitor and counter evolving threats, and vastly speed up digital forensics techniques. Add to this the increased risk of adversarial attacks on machine learning, deep learning, and autonomous systems, and you have a field that’s poised to grow massively over the next decade. This is a new focus area for ODSC. West features some of the leading experts in the CyberML field.
Register Now for ODSC West 2021
At our upcoming event this November 16th-18th in San Francisco, ODSC West 2021 will feature a plethora of talks, workshops, and training sessions on machine learning topics, deep learning, NLP, MLOps, and so on. You can register now for 20% off all ticket types, or register for a free AI Expo Pass to see what some big names in AI are doing now. Some highlighted sessions on machine learning topics include:
- Towards More Energy-Efficient Neural Networks? Use Your Brain!: Olaf de Leeuw | Data Scientist | Dataworkz
- Practical MLOps: Automation Journey: Evgenii Vinogradov, PhD | Head of DHW Development | YooMoney
- Applications of Modern Survival Modeling with Python: Brian Kent, PhD | Data Scientist | Founder The Crosstab Kite
- Using Change Detection Algorithms for Detecting Anomalous Behavior in Large Systems: Veena Mendiratta, PhD | Adjunct Faculty, Network Reliability and Analytics Researcher | Northwestern University
Sessions on MLOps:
- Tuning Hyperparameters with Reproducible Experiments: Milecia McGregor | Senior Software Engineer | Iterative
- MLOps… From Model to Production: Filipa Peleja, PhD | Lead Data Scientist | Levi Strauss & Co
- Operationalization of Models Developed and Deployed in Heterogeneous Platforms: Sourav Mazumder | Data Scientist, Thought Leader, AI & ML Operationalization Leader | IBM
- Develop and Deploy a Machine Learning Pipeline in 45 Minutes with Ploomber: Eduardo Blancas | Data Scientist | Fidelity Investments
Sessions on Deep Learning:
- GANs: Theory and Practice, Image Synthesis With GANs Using TensorFlow: Ajay Baranwal | Center Director | Center for Deep Learning in Electronic Manufacturing, Inc
- Machine Learning With Graphs: Going Beyond Tabular Data: Dr. Clair J. Sullivan | Data Science Advocate | Neo4j
- Deep Dive into Reinforcement Learning with PPO using TF-Agents & TensorFlow 2.0: Oliver Zeigermann | Software Developer | embarc Software Consulting GmbH
- Get Started with Time-Series Forecasting using the Google Cloud AI Platform: Karl Weinmeister | Developer Relations Engineering Manager | Google