CSCE 479/879 (Fall 2019) Project Rules and Ideas

From the syllabus:

In this course you and your team will do a substantial project, in which you will characterize a significant problem amenable to a deep learning solution, study the related work to this problem, develop one or more deep learning approaches to this problem, and evaluate your approaches.

You will summarize your project results in a written report and an oral presentation. The written report must use a professional writing style similar to that found in a refereed conference or journal (e.g., ACM, IEEE, ICML, ICLR, NeurIPS), including abstract, introduction, summary of related work, your contribution, references, and an appendix (if necessary). The oral presentation will be to the entire class at the end of the semester: during the fifteenth week (December 9–13), and if necessary, during the fourteenth week (December 2–6). You will submit your written report no later than 11:59 p.m. on Sunday, December 15. In accordance with UNL policies, you have now been informed in writing of the nature and scope of this project prior to the eighth week of classes.

Seven to eight weeks into the semester (early to mid-October), we will set a deadline for submission of proposals on your projects. You must do this in order to get full credit for your project, and you must get our approval on your proposal before starting work on your project. A couple of weeks later, you will submit to us an initial progress report and meet with us for a check-in on your project. For this progress report, you will have your data sets downloaded and preprocessed, and have all necessary libraries and environments installed (or at least have the installation requests pending with HCC). A few weeks after that, you will submit to us your second progress report. This report will include initial experimental results and will serve as an early draft of your progress report, so it will also include your problem definition, related work, and discussion on your approaches.

Projects are officially due by 11:59 pm on Sunday, December 15. See the rules on projects for more information.

Project oral presentations are the weeks of December 2 and December 9. See the presentation schedule for more information.

✔ Second project progress reports are due Sunday, November 17. You will submit it in pdf format via handin Canvas. The report, which is considered an early draft of your final report, should include:

✔ First project progress reports are due Sunday, November 3. You should submit it to sscott@cse and equint@cse in text format in the body of an email via Canvas before 11:59 pm on that day. The report should include:

Project Ideas

The following is a list of possible projects, suggested by a variety of individuals. If you want to know more about a particular one, let us know and we can put you in contact with those who can provide more details.

You are welcome to suggest your own project as well, e.g., one related to your own research. It should contain some form of an experimental component, be relevant to the course, and be a non-trivial extension beyond what we covered in class.

Classification

• Reproducibility Challenge
• Problem: One of the challenges in machine learning research is to ensure that published results are reliable and reproducible. In support of this, the objective of the reproducibility challenge is to investigate reproducibility of papers appearing in prominent conferences. The most recent challenge is to reproduce results of papers submitted to the 2019 Neural Information Processing Systems (NeurIPS) conference. The challenge consists of choosing an accepted paper and reproducing its results. The challenge's submission deadline is December 1, so there is time to work on reproducing results for a course project and submit to the challenge.
• Reproducibility Challenge @ NeurIPS 2019


• Generating item names from online fantasy games
• Problem: Given a dataset containing approximately 75,000 names from various fantasy-style Massive Multiplayer Online Role-Playing Games (MMORPGs) such as World of Warcraft, among others, generate item names. This natural language processing problem would likely use a recurrent neural network architecture.


• Code recommendation
• Problem: Apply graph convolutional networks to code recommendation systems such as Aroma (citations below). Each method can be represented as an abstract syntax tree (AST) and labels for nodes in AST would be the method name. There are large open source projects available in repository management services like GitHub and GitLab for corpus for a given language.
• Aroma: Using machine learning for code recommendation
• Aroma: Code Recommendation via Structural Code Search


• Deep learning in wireless networks
• Problem: Apply deep learning to problems in wireless networking. Data collected from the NEXTT testbed of a cloud-radio access network (C-RAN)-based experimental wireless network. Possibilities include, but are not limited to, channel estimation, bandwidth sharing, and others related to massive MIMO. Could become a thesis and a paper.


• Defense against adversarial examples
• Problem: Several image classification systems have been shown to be tricked into gross misclassification by making small, sometimes imperceptible changes to the input image, e.g., tricking the system into thinking a panda is a gibbon. An open problem is how to modify an architecture or regularizer to make classifiers more robust against such minor changes.
• Related papers:
• Intriguing properties of neural networks
• Explaining and Harnessing Adversarial Examples
• One pixel attack for fooling deep neural networks
• Adversarial Patch
• Characterizing Adversarial Subspaces Using Local Intrinsic Dimensionality
• Certifiable Distributional Robustness with Principled Adversarial Training

Reinforcement Learning

• Learning how to apply genetic algorithm operators for software assurance
• Problem: Use reinforcement learning to learn a policy to select which genetic algorithm (GA) action is most appropriate for evolving solutions for formal methods of software assurance.
• Data: Kodkod relational models are fed into a GA for solving. The RL agent will learn how to select which GA operators for fastest solving.
• Proposer: Hamid Bagheri
• Related Paper:
• Selecting evolutionary operators using reinforcement learning: Initial explorations


• Play a new game with AlphaZero
• Problem: Re-implement Deepmind's famous approach to Go/Chess/Shogi for a new game
• Data: Environment of students' choosing, ideally some difficult, perfect knowledge, deterministic, two player game
• Proposer: Eleanor Quint, with willingness to collaborate after the semester
• Related papers and blog posts:
• AlphaGo Zero: Learning from scratch
• Mastering the game of Go with deep neural networks and tree search
• Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm


• Play Starcraft 2 Minigames
• Problem: Play minigames in the StarCraft II Learning Environment, a cutting edge problem in Deep RL
• Environment: https://github.com/deepmind/pysc2
• Notes: This project could go in a lot if directions involving planning, AI, multi-agent systems, and more
• Proposer: Eleanor Quint, with willingness to collaborate after the semester
• Related papers and blog posts:
• DeepMind and Blizzard open StarCraft II as an AI research environment
• Population Based Training of Neural Networks (possibly state of the art on some SC2 tasks)
• MSC: A Dataset for Macro-Management in StarCraft II

Imputation

• Predicting missing flight recorder data
• Problem: Given data (time series or fixed window) from an airplane's flight recorder, predict values that are missing in the sequence. Could be a thesis or a paper.

Autoencoding

• Time-varying probability density estimation
• Problem: Modeling probabilty densities is fundamental to many scientific areas, including ecology and molecular dynamics. Often, such density models evolve in time. E.g., in molecular dynamics, at time t=0 the density of atoms in a 3D space might be near uniform, but over time it becomes progressively less so, peaking at locations where atoms are more likely to appear, due to self-assembly of molecular structures.
  This project's goal is a feasibility study into learning deep generative models that are time-sensitive. Variational autoencoders, generative adversarial networks, and normalizing flows are examples of deep generative models that might be amenable to adding a time dimension, perhaps adding time as an additional input (or output) and regularizing to enforce that the model evolves smoothly in time. Experiments can involve simple 1D or 2D synthetic data, as well as 3D data from molecular dynamics studies.


• Hard instance generation
• Problem: Some computational problems are known to be intractable in the worst case, but are still widely studied at a heuristic level. One such problem is 3-CNF-SAT: given a boolean 3-CNF formula φ, answer whether there an assignment to its variables that satisfies it. heuristics for 3-CNF-SAT have been studied extensively, but there are still some formulas that are very difficult to solve (requiring days of processing). This project's goal is to apply autoencoders to characterize these hard problems, to lend insight on why hard instances are hard.
• Data from competitions


• Action-Conditional Video Prediction
• Problem: Create a model which, conditioned on the current observation of a stochastic MDP environment (e.g., Atari or robotic pushing environment) and the actions taken by an existing policy, predicts future states of the stochastic environment. Very strongly related to reinforcement learning theory and applications
• Proposer: Eleanor Quint, with high willingness in long-term collaboration after the semester and to push ideas to publication
• Related papers and blog posts:
• Action-Conditional Video Prediction using Deep Networks in Atari Games Blog Post
• Unsupervised Learning for Physical Interaction through Video Prediction
• Deep Visual Foresight for Planning Robot Motion
• Stochastic Variational Video Prediction

Miscellaneous Applications

• An application of graph convolutional kernels
• Problem: Graph convolutional kernels are analogous to convolutions on images, but they try to identify features in the adjacency matrices of graphs. Useful in node labeling, edge identification, and graph compression.
• Data: Graph data from the Stanford Large Network Dataset Collection (SNAP), and other graphs
• Related papers and blog posts:
• GRAPH CONVOLUTIONAL NETWORKS blog post


• Application of convolutional neural networks for genomic motif finding
• Problem: CNNs have been applied to natural language processing (NLP). The phrases are encoded into a format that looks like Braille and input to the network. This project is to determine if that could be used for finding genomic motifs for bacterial species. Conventionally, these motifs are k-mers and their frequencies in a species' genome is considered a signature for that species. Perhaps one could use a CNN and autoencoder to come up with relevant motifs instead of enumerating all k-mers. The data in this case would be sequencing reads.
• Related links:
• Text Understanding from Scratch
• Sequencing data



---

Last modified 20 November 2019; please report problems to sscott.