Further Reading#
These notes cite the core sources used throughout the course, but students often benefit from a small amount of reading guidance rather than a flat bibliography. A good way to use the references is to distinguish between background texts, canonical first papers for each model family, and more advanced follow-up papers that reveal how the field evolved.
For broad background, [Bis06] remains a strong source for probabilistic modeling language, while [GBC16] is still a useful general deep-learning reference for neural architectures, optimization, and representation learning. Students who feel less comfortable with the foundations chapters should begin there before diving more deeply into the specialized generative-model papers.
Suggested Reading Order#
A sensible reading path after the course is the following.
First, for latent-variable generative modeling, read the original VAE papers [KW13] and [RMW14]. They are concise and reveal the core variational logic that later reappears in more sophisticated models. Next, for adversarial learning, read [GPAM+14], then compare it with stabilization-oriented work such as [ACB17], [MKKY18], and [ZPIE17] to see how quickly the field moved from a simple game formulation to a broader family of objectives and constraints.
For diffusion, a good route is to read [HJA20] first for the discrete DDPM perspective, then [SSDK+21] for the continuous-time score-based view, and then [SME21] for deterministic sampling. Students interested in modern large-scale systems should follow these with [RBL+22], which shows how latent-space diffusion made high-resolution synthesis far more practical.
Reading by Chapter Theme#
The chapters of the book line up naturally with particular papers. The probabilistic latent-variable thread is anchored by [KW13] and [RMW14]. The adversarial thread begins with [GPAM+14] and branches into discrepancy design and stabilization via [ACB17] and [MKKY18]. The stochastic transport and score-based thread is represented by [HJA20], [SE20], [SSDK+21], and [SDME21]. The deterministic transport and flow-matching thread is represented by [LCBH+23] and more recent path-design work such as [TFM+24].
A useful meta-question when reading any of these papers is the same one that structures the course itself: what is the chosen source of randomness, what object is the neural network asked to predict or represent, and what surrogate objective makes learning feasible? Asking those three questions repeatedly makes the literature much easier to organize.
How To Read the Papers#
Students reading these references for the first time should not expect every proof or implementation detail to be equally central. A productive first pass is usually:
identify the probabilistic objects being modeled;
identify the trainable neural components;
identify the obstacle to direct maximum-likelihood learning;
identify the surrogate objective or transport reformulation that resolves that obstacle.
Once that scaffold is clear, the paper’s specific derivations and engineering choices become much easier to interpret. This is especially important for diffusion and flow-matching papers, where the notation can look heavy before the conceptual structure is visible.
Martin Arjovsky, Soumith Chintala, and Léon Bottou. Wasserstein generative adversarial networks. Proceedings of the 34th International Conference on Machine Learning, 2017.
Andrea Asperti, Davide Evangelista, and Elena Loli Piccolomini. A survey on variational autoencoders from a green ai perspective. SN Computer Science, 2021. arXiv:2103.01071, doi:10.1007/s42979-021-00702-9.
Christopher M. Bishop. Pattern Recognition and Machine Learning. Springer, 2006.
Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2019.
Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep Learning. MIT Press, 2016.
Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Advances in Neural Information Processing Systems. 2014.
Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, Vincent Dumoulin, and Aaron Courville. Improved training of wasserstein gans. In Advances in Neural Information Processing Systems, volume 30. 2017.
Irina Higgins, Loic Matthey, Arka Pal, Christopher Burgess, Xavier Glorot, Matthew Botvinick, Shakir Mohamed, and Alexander Lerchner. Beta-vae: learning basic visual concepts with a constrained variational framework. In International Conference on Learning Representations. 2017.
Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems, 2020.
Diederik P. Kingma and Max Welling. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114, 2013.
Yaron Lipman, Ricky T. Q. Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. International Conference on Learning Representations, 2023.
Mehdi Mirza and Simon Osindero. Conditional generative adversarial nets. arXiv preprint arXiv:1411.1784, 2014.
Takeru Miyato, Toshiki Kataoka, Masanori Koyama, and Yuichi Yoshida. Spectral normalization for generative adversarial networks. International Conference on Learning Representations, 2018.
Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. In Proceedings of the 38th International Conference on Machine Learning. 2021.
Danilo Jimenez Rezende, Shakir Mohamed, and Daan Wierstra. Stochastic backpropagation and approximate inference in deep generative models. In Proceedings of the 31st International Conference on Machine Learning. 2014.
Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022.
Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. Proceedings of the 32nd International Conference on Machine Learning, 2015.
Kihyuk Sohn, Honglak Lee, and Xinchen Yan. Learning structured output representation using deep conditional generative models. In Advances in Neural Information Processing Systems, volume 28. 2015.
Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. International Conference on Learning Representations, 2021.
Yang Song, Conor Durkan, Iain Murray, and Stefano Ermon. Maximum likelihood training of score-based diffusion models. Advances in Neural Information Processing Systems, 2021.
Yang Song and Stefano Ermon. Denoising score matching for generative modeling. Advances in Neural Information Processing Systems, 2020.
Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. International Conference on Learning Representations, 2021.
Alexander Tong, Kilian Fatras, Grégoire Mialon, Yann LeCun, Nicolas Courty, and Rémi Flamary. Improving and generalizing flow-based generative models with minibatch optimal transport. Transactions on Machine Learning Research, 2024.
Aaron van den Oord, Oriol Vinyals, and Koray Kavukcuoglu. Neural discrete representation learning. In Advances in Neural Information Processing Systems, volume 30. 2017. arXiv:1711.00937.
Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A. Efros. Unpaired image-to-image translation using cycle-consistent adversarial networks. In Proceedings of the IEEE International Conference on Computer Vision. 2017.