We’re again from the NeurIPS 2023 convention. It was really inspiring to interact with the AI group and be taught extra in regards to the present developments and way forward for AI. We additionally loved the chance to share how you should use MATLAB to combine AI into cyber-physical techniques. On this weblog put up, I’m going to current some highlights from the MATLAB AI group’s presence on the convention.
Determine: Our group on the MathWorks sales space
Left to proper (entrance): Ashvant Ram Selvam, Lucas García, Brenda Zhuang, Sivylla Paraskevopoulou, Anoush Najarian, Maitreyi Chitale, and Jon Cherrie
Left to proper (again): Mary Ann Freeman, Antoni Woss, Naren Srivaths Raman, Conor Daly, Jianghao Wang, Kevin Holly, Elvira Osuna Highley, and Abhijit Bhattacharjee
Determine: Lucas García and Naren Srivaths Raman presenting talks on Incorporating ML Fashions into Security-Essential Techniques and Reinforcement Studying: Developments, Functions, and Challenges, and Maitreyi Chitale presenting a poster on Farm-To-Plate-AI.
Determine: (left) The robotic arm and Naren, the robotic programmer. (proper) Elvira taking part in Tic-Tac-Toe in opposition to the robotic.
Watch the video under to see the robotic in motion.
Determine: On the WiML (Ladies in Machine Studying) occasion at NeurIPS.
Determine: Our group on the MathWorks sales space
Left to proper (entrance): Ashvant Ram Selvam, Lucas García, Brenda Zhuang, Sivylla Paraskevopoulou, Anoush Najarian, Maitreyi Chitale, and Jon Cherrie
Left to proper (again): Mary Ann Freeman, Antoni Woss, Naren Srivaths Raman, Conor Daly, Jianghao Wang, Kevin Holly, Elvira Osuna Highley, and Abhijit Bhattacharjee
3 Talks and A Poster
Right here, yow will discover an summary of the talks and poster introduced by MathWorks at NeurIPS, together with hyperlinks to slides.
Determine: Lucas García and Naren Srivaths Raman presenting talks on Incorporating ML Fashions into Security-Essential Techniques and Reinforcement Studying: Developments, Functions, and Challenges, and Maitreyi Chitale presenting a poster on Farm-To-Plate-AI.
Speak 1
From Idea to Apply: Incorporating ML Fashions into Security-Essential Techniques
Lucas García On this discuss, Lucas delved into the complexities of certifying neural networks for safety-critical purposes, with a selected concentrate on the aviation {industry}. The presentation launched AI Certification within the aviation {industry}, and a case examine for a runway signal classification system within the cockpit designed to enhance pilot navigation and situational consciousness. By strolling the viewers by way of every step of the workflow—from requirement definition to mannequin coaching and system validation—the discuss supplied a complete blueprint for integrating neural networks into safety-critical techniques for low criticality purposes, paving the way in which for future developments in industry-specific certification requirements. SlidesSpeak 2
Utilizing AI to Enhance Management Design Workflows
Naren Srivaths Raman Management techniques are ubiquitous and allow the secure and predictable operation of airplanes, vehicles, and power techniques. Simply as in different engineering disciplines, management engineers have an interest within the new prospects AI affords to reinforce conventional options. The discuss lined a number of areas the place AI is gaining curiosity and adoption amongst management engineers and researchers. On this discuss, Naren (1) explored AI for modeling a system managed (plant) with strategies reminiscent of nonlinear system identification and lowered order modeling (ROM), (2) mentioned the usage of AI for digital sensor modeling and management algorithm design, particularly, the design of nonlinear mannequin predictive management (MPC) utilizing neural state-space (NSS) fashions, and (3) touched upon how reinforcement studying (RL) can be utilized as a software for controller tuning or can substitute a conventional controller altogether. Lastly, Naren launched some greatest practices in utilizing AI for controls to deal with challenges, reminiscent of lack of efficiency ensures. SlidesSpeak 3
Reinforcement Studying: Developments, Functions, and Challenges
Naren Srivaths Raman Reinforcement studying (RL) has been gaining consideration as a machine studying approach that may routinely be taught advanced behaviors and understand excessive efficiency. RL purposes span varied domains, together with management design, robotics, automated driving, communications, and extra. Nonetheless, reinforcement studying comes with a number of challenges. These embody the necessity for giant quantities of coaching knowledge, difficulties in tuning hyperparameters, and verification of deep neural community insurance policies. On this discuss, Naren mentioned tendencies, purposes, and challenges, and launched concepts, instruments, and greatest practices on find out how to tackle these challenges. SlidesPoster
Farm-to-Plate-AI: Enhancing Freshness and Decreasing Waste utilizing Laptop Imaginative and prescient and Robotics
Maitreyi Chitale, Brenda Zhuang, et al. Addressing international meals safety and decreasing waste within the meals provide chain are essential challenges in right now’s world. The poster highlights the transformative potential of integrating rising applied sciences, reminiscent of autonomous navigation, synthetic intelligence (AI), and pc imaginative and prescient, to optimize farming practices, automate harvesting and grading, and monitor meals high quality throughout transportation. By leveraging these applied sciences, we are able to successfully tackle the challenges posed by conflicts, pandemics, local weather change, inequality, and the projected 70% improve in meals demand as a consequence of inhabitants development. We constructed a workshop with workouts to simulate the pipeline, which included simulating a UAV by way of completely different farm configurations, using YOLOv4 and performing switch studying on it with a smaller dataset to carry out object counting of mangoes pre-harvest, and using machine studying on hyperspectral photographs to gauge the ripeness of mangoes. The workshop was introduced at GHC23 to 750+ in-person individuals, in addition to at digital WiDS workshop collection and past. Our poster, which is predicated on the workshop we created, proposes an end-to-end pipeline, from farm to plate and leverages the mixed positive aspects of particular person AI and robotics algorithms to optimize the worldwide meals provide chain and obtain vital waste discount.Demos
As promised (in a earlier weblog put up) , we had a number of goodies at our each and funky demos displaying the AI Chat Playground (the brand new GenAI software for producing MATLAB code), real-time occasion segmentation on a GPU, and the group favourite demo – a robotic taking part in Tic-Tac-Toe. An RL agent was educated to play the sport and the robotic used pc imaginative and prescient for recognizing the progress of the Tic-Tac-Toe recreation on the board.
Determine: (left) The robotic arm and Naren, the robotic programmer. (proper) Elvira taking part in Tic-Tac-Toe in opposition to the robotic.
Watch the video under to see the robotic in motion.
Video (4x): Robotic utilizing reinforcement studying and pc imaginative and prescient is taking part in Tic-Tac-Toe.
See You Subsequent Yr
We’re trying ahead to becoming a member of NeurIPS once more subsequent yr for extra sharing, occasions, and sales space interactions. Till then, go away your remark right here to speak about AI, your workflows, and challenges.
Determine: On the WiML (Ladies in Machine Studying) occasion at NeurIPS.
