Subgroup Meeting Materials

https://docs.google.com/document/d/15CBu4rlj8MSlLospoAaMnLtjQ3dri74QCqSOVQlQ6N4/edit?usp=sharing

May 29, 2024 10:00 AM:

Research Activity: (Brian Zambrano)

I will review a CNN architecture and methodology to train in identifying methane plumes on satellite images delivered by Sentinel-5p satellite. I will also give a general idea of how satellite images are used to retrieve methane concentration and methane plumes and how these ideas can be applied to other satellite images, especially Sentinel-2. 

Writing-Reading Activity/ Current Trends in AI/M: (Esha Saha)

I will discuss some tips on how to give an effective presentation. They will emphasize on some useful pointers that most of us know but often miss out while actually giving a presentation.

June 5, 2024:

Research Activity: (Shan Gao)

Mountain pine beetle (MPB) infestation has been a long-lasting problem, resulting in boom-bust cycles of trees every two to four decades. I am going to introduce how the MPB attack dynamics and Lodgepole pine dynamics were modeled.

Writing-Reading Activity/ Current Trends in AI/M: (Miles Kent)

I will discuss model and data parallelism, and why they are often used in today's machine learning modeling and data science tasks. In addition, I will also share some resources in both R and Python that will allow you to get started implementing both in your research.

June 12, 2024:

Research Activity: (Binbing Wu)

I will discuss Predicting discrete-time bifurcations with deep learning. The same method is used for continuous time bifurcation (LSTM-CNN). Discrete-time bifurcation is used to produce training data by five different models.

Writing-Reading Activity/ Current Trends in AI/M: (Brian Zambrano)

I will discuss the documentation, doc strings, and comments in the code presentation. Following the increase in code use in mathematical papers, journals are starting to require the presentation of the code as complementary material. For this reason, it is a good practice to include doc string and comets that make the code easier to understand.

June 19, 2024:

Research Activity: (Dongxuan Li)

In this paper, we investigate the generalization capabilities of classifiers trained on the ImageNet dataset. Despite achieving high accuracy on the original ImageNet validation set, the robustness of these models in practical, real-world scenarios remains uncertain.

Writing-Reading Activity/ Current Trends in AI/M: (Maulik Srivastava)

I will showcase a unique neural network architecture called MobileNet and discuss what makes it inherently suitable for environments with limited computational resources.

June 26, 2024:

Research Activity: (Anuththara Sarathchandra)

I will present about the Linear Threshold Model which can be used to model some collective behaviors of individuals in a population and the threshold identifiability of the Linear Threshold model.

Writing-Reading Activity/ Current Trends in AI/M: (Binbing Wu)

Stable diffusion generate images(QR code)

July 3, 2024: NO MEETING

July 10, 2024:

Research Activity: (Oscar Wang)

I will give a brief description on the paper: "Second-generation stoichiometric mathematical model to predict methane emissions from oil sands tailings" and talk about how this is helping our research on building a model with chemical and weather factors influencing the emission of methane in the oiling industry.

Writing-Reading Activity/ Current Trends in AI/M: (Dongxuan Li)

This paper mainly focused on two methods: the Monte Carlo search tree and divide and conquer to find the best path from the initial points to the destination. A brief summary could appropriately describe the real situation. Not only the path we need to search using the divide and conquer, but the direction also has multiple choices so normally using the Monte Carlo search tree to distinguish the best direction. We can discuss more details and examples during the meeting.

July 17, 2024:

Research Activity: (Esha Saha)

I will be presenting the math behind diffusion models, a popular technique in generative AI. The presentation will cover details of how a diffusion model is derived mathematically along with its training procedure. Further, time permitting I will give some results from my own research on developing interpretable diffusion models using random feature models that help us to theoretically bound the distance between generated and true input distribution of data.

Writing-Reading Activity/ Current Trends in AI/M: (Shan Gao)

This is the continuation of my previous talk. I will introduce the basic form of the mountain pine beetle dynamic. Then I will present several classic models and update my progress.

Jul 24, 2024:

Research Activity: (Maulik Srivastava)

I will be describing my ongoing research on the prediction of wildfires using machine learning and satellite image data, with a focus on data collection/pre-processing techniques and an overview of the model architecture.

Writing-Reading Activity/ Current Trends in AI/M: (Amit Chakraborty)

The main idea of this activity was to learn Bayesian network from data using bnlearn R package and make predictions with it. I will be describing why the Bayesian network is useful and how it relates to conditional independence. I will start with the importance of joint distribution and then conditional independence and then Bayesian networks. Finally, I will show a R code about how to utilize the bnlearn package to learn the network using different learning algorithms and how to make predictions from it. 

July 31, 2024:

Research Activity: (Miles Kent)

I will be describing Binbing and I’s ongoing research project, which involves developing a tool to assess climate related physical risk of heat islands and floods. I will first give an overview of the project and what climate related physical risks are. This will then be followed by the project goals and a brief discussion of how we plan to attack each part of our research problem 

Writing-Reading Activity/ Current Trends in AI/M: (Oscar Wang)

Brief introduction to Multimodel AI, and how it works basically.

Schedule for the 2023 Summer

July 5: Ilhem

July 12: Shan

July 19: Russell

July 26: Miles

August 2: Amit

• Date: Aug 2, 2023

• Participants: Russell, Jacob, Miles, Amit, Ruoshi, Rong, Dongxuan, Shan, Binbing

• Activities: Amit gave a presentation showing an update on his summer project. The presentation covered the problem review, training data generation methods, and the latest experimental results.

• Date: Jul 26, 2023

• Participants: Russell, Jacob, Miles, Amit, Ruoshi, Rong, Dongxuan, Shan, Binbing

• Activities: Miles gave a presentation explaining how to use cluster computing resources from Compute Canada.

• Date: Jul 5, 2023

• Participants: Russell, Ilhem, Amit, Jacob, Miles, Rong, Dongxuan, Shan, Binbing

• Activities: Shan gave the presentation. Shan provided an introduction to the Early Warning System (EWS) approach and its associated statistical concepts. He elaborated on the methodology employed for generating the training data. Additionally, he presented the results derived from employing statistical techniques.

• Date: Jul 5, 2023

• Participants: Russell, Ilhem, Jacob, Miles, Ruoshi, Rong, Dongxuan, Shan

• Activities: Ilhem gave the presentation. Here is the abstract: the cytosol's spatial structure and physical properties are important for many cellular processes, including molecular transport, signalling, and metabolic reactions. However, due to the complexity and dynamic nature of the cytosol, its spatial structure and physical properties still need to be better understood. We aim in this work to estimate the heterogeneity of the diffusion coefficient across the cytosol to emphasize how crowding may impact the formation of biomolecular condensates within cells. Our approach enables us at once to perform particle tracking and estimate particle diffusion through Bayesian inference. The methodology developed could be implemented to analyze the functional compartmentalization of the cytosol for the nuclear division cycle within the hypha of multinucleate fungi.

• Date: Nov 3, 2022

• Participants: Russell Milne, Pijush Panday, Amit Chakraborty, Lin Wang, Jianzhong Gao, Xiaoqi Xie

• Activities: Xiaoqi presented her work on predicting mountain pine beetle outbreaks using statistical methods. She then asked some questions to the group members about her doubts. Russell gave her some suggestions. Finally, Russell led a discussion on how to improve scientific writing.

• Date: Oct 27, 2022

• Participants: Russell Milne, Amit Chakraborty, Lin Wang, Jianzhong Gao, Pijush Panday, Xiaoqi Xie

• Activities: Lin briefly discussed her work on disease elimination in epidemic models with lockdown policy. She showed that in her model, EWS indicators was not showing proper trends in the approach to disease elimination. Other group members asked questions and gave some suggestions to her. Amit also had some queries about his work regarding the inverse method, which he asked during the research discussion. Other members gave some valuable tips.

• Date: Oct 20, 2022

• Participants: Amit Chakraborty, Lin Wang, Jianzhong Gao, Pijush Panday

• Activities: Pijush discussed bifurcation theory and the occurrence of different kinds of bifurcation in epidemic models. Then he briefly discusses his current work of predicting the disease emergence using deep learning model. Finally, there was a discussion among all members of the subgroup.

• Date: Oct 13, 2022

• Participants: Pijush Panday, Amit Chakraborty, Lin Wang, Jianzhong Gao, Xiaoqi Xie

• Activities: Xiaoqi briefly discussed her current work on mountain pine beetle dynamics using statistical modelling. Amit gave a brief overview of his work on using GBM and Bayesian network approaches to forecast the COVID-19 cases in USA. Finally, other members asked some questions to them.