Foundations of Machine Learning Boot Camp 2017
2017-01-23..2017-01-30
Berkeley, United States
[Foundations of Data Science Boot Camp 2018, Frontiers of Deep Learning 2019, Data-Driven Decision Processes Boot Camp 2022, Graph Learning Meets Theoretical Computer Science 2025, Bridging Continuous and Discrete Optimization Boot Camp 2017, Emerging Challenges in Deep Learning (Simons Institute workshop, 2019)]
who it's for
machine learning researchers, graduate students in machine learning, bayesian statisticians, reinforcement learning researchers, deep learning practitioners, robotics engineers, data scientists working with probabilistic models
topics
stochastic gradient descent and AdaGrad; variance reduction methods (SVRG); Frank-Wolfe projection-free optimization; submodular optimization and Lovász extension; continuous greedy multilinear relaxation; Markov decision processes and Thompson sampling; deep reinforcement learning (DQN, policy/value nets); nonparametric Bayesian stick-breaking models; Pitman-Yor and completely random measures; tensor decompositions for latent variables; imitation learning (DAgger) and guided policy search
| title | author | length |
|---|---|---|
| Optimization for Machine Learning I | Elad Hazan | 01h 05m |
Casts learning as optimization over parameters and proves convex relaxations let local gradient info give global guarantees; uses stochastic gradient descent… | ||
| Optimization for Machine Learning II | Elad Hazan | 01h 04m |
AdaGrad is presented as the adaptive regularizer used in practice for deep nets; stochastic gradient descent needs O(1/epsilon^2) iterations while smooth full… | ||
| Submodularity: Theory and Applications I | Stefanie Jegelka | 01h 04m |
Defines submodularity as diminishing marginal gains and claims a practical recipe: convert a set function to the Lovász extension (a convex, piecewise‑linear… | ||
| Submodularity: Theory and Applications II | Andreas Krause | 01h 04m |
Greedy selection provably achieves a 1-1/e ≈ 0.63 approximation for monotone submodular maximization under a cardinality constraint and is optimal with oracle… | ||
| A Tutorial on Reinforcement Learning I | Emma Brunskill | 01h 01m |
Concise tutorial on reinforcement learning that defines MDPs, contrasts model‑based vs model‑free (tabular Bellman backups and Q‑learning), and treats… | ||
| A Tutorial on Reinforcement Learning II | Emma Brunskill | 56m |
Summarizes regret bounds and practical fixes: UCRL-2 gives regret scaling O(D · S · sqrt(A T) · log(T/δ)) under finite diameter while gap-dependent bounds can… | ||
| Interactive Learning of Classifiers and Other Structures | Sanjoy Dasgupta, Rob Nowak | 01h 31m |
Interactive learning studies when an algorithm queries labels or other feedback adaptively and what guarantees that yields: active learning can cut label needs… | ||
| Deep Robotic Learning | Sergey Levine | 01h 35m |
Argues end‑to‑end sensorimotor learning beats hand‑engineered pipelines by closing perception and control, shows concrete algorithms and tradeoffs: DAgger… | ||
| Nonparametric Bayesian Methods: Models, Algorithms, and Applications I | Tamara Broderick | 01h 06m |
Argument: treat the latent mixture as an infinite parameter so realized cluster count grows with data (roughly log N in the basic Dirichlet case)… | ||
| Nonparametric Bayesian Methods: Models, Algorithms, and Applications II | Michael Jordan | 01h 04m |
Covers how stick‑breaking GEM priors produce discrete random measures used as Dirichlet process priors, derives the Chinese restaurant and Pólya‑urn… | ||
| Nonparametric Bayesian Methods: Models, Algorithms, and Applications III | Michael Jordan | 01h 02m |
Shows how stick-breaking and the Polya/Chinese Restaurant logic arise from a Beta–Bernoulli De Finetti mix, then constructs completely random measures by… | ||
| Nonparametric Bayesian Methods: Models, Algorithms, and Applications IV | Tamara Broderick | 01h 04m |
Explains why node exchangeability (Aldous–Hoover/graphon representation) forces dense or empty growing graphs with edges ~N^2, and proposes edge… | ||
| Tutorial on Deep Learning I | Ruslan Salakhutdinov | 01h 02m |
Claims that modern deep nets win only with very large datasets and practical tricks: hierarchical representations work well for vision, language and even… | ||
| Tutorial on Deep Learning II | Ruslan Salakhutdinov | 56m |
Claims concrete wins: AlexNet halved error on ImageNet (1.2M images), 8 layers, ~50M parameters, ~18% top‑5 and removing one FC layer drops only ~1% while… | ||
| Tutorial on Deep Learning III | Ruslan Salakhutdinov | 57m |
Compares undirected deep Boltzmann machines and directed variational autoencoders, explains why mean-field approximation gives fast, unimodal posterior… | ||
| Tutorial on Deep Learning IV | Ruslan Salakhutdinov | 01h 00m |
Covers how GANs as a min/max game learn a generator and differentiable discriminator that sample (no explicit density), showing deconvolution GAN images on… | ||
| Tensor Decompositions for Learning Latent Variable Models I | Daniel Hsu | 56m |
Claims that a mixture of multinomial topic models with one topic per document is identifiable from third-order moments when topic word-vectors are linearly… | ||
| Tensor Decompositions for Learning Latent Variable Models II | Daniel Hsu | 53m |
Shows that third-order moments plus a whitening reduction recover latent mixtures when component means are linearly independent and weights are bounded away… | ||
| Natural Language Understanding: Foundations and State-of-the-Art | Percy Liang | 01h 31m |
Survey of three concrete semantics approaches: distributional (Word2Vec / LSA) for broad coverage, frame semantics to expose argument structure, and… | ||
| nothing matches | ||