Alex Anderson

Hello! My name is Alex and I am a Ph.D. Candidate at UC Berkeley working with Bruno Olshausen in the Redwood Center for Theoretical Neuroscience. I will be starting as a Research Scientist at WaveOne in July 2018. I am funded through the NSF GRFP and Berkeley Deep Drive. I use machine learning to understand human vision and organize events to improve my community.

In the summer of 2016, I interned at Oculus Research. In the past, I graduated summa cum laude at Washington University in St. Louis with a double major in math and physics. I did my undergraduate thesis with Carl Bender. Before that, I enjoyed high school math competitions.

LinkedIn | GitHub | Email | Resume

Projects

Journal Article: High-Acuity Vision from Retinal Image Motion
Alex Anderson, Kavitha Ratnam, Austin Roorda, Bruno Olshausen (2018)
A mathematical model shows how the eyes self-generated drift motion can improve high-acuity vision by averaging over retinal inhomogeneities. Under Review
Conference Paper: The High Dimensional Geometry of Binary Neural Networks
Alexander G. Anderson, Cory P. Berg International Conference on Learning Representations (ICLR) Conference Track (April 30th 2018)
Recent successes of Binary Neural Networks can be understood based on the geometry of high-dimensional binary vectors. Paper | Code
Conference Paper: Joint Source-Channel Coding with Neural Networks for Analog Data Compression and Storage in Emerging Devices.
Ryan Zarcone, Dylan Paiton, Alexander G. Anderson, Jesse Engel, H.S. Phillip Wong, Bruno A. Olshausen. Data Compression Conference (March 27th 2018)
Neural networks are used to store images in noisy memory devices.
Poster Presentation: The High Dimensional Geometry of Binary Neural Networks
Alexander G. Anderson, Cory P. Berg International Conference on Learning Representations (ICLR) Workshop Track (April 24th-26th 2017)
Due to the high dimensional geometry of binary vectors, continuous feature vectors can be well-approximated by binary vectors. This gives a theoretical foundation for the success of the Courbariaux, Hubara, et al. method of training neural networks with binary weights and activations. The theory is also validated in such networks. Paper and Review
Summer Internship: Predictive Eye Tracking Using Recurrent Neural Networks
Alexander G. Anderson, Alexander Fix, Rob Cavin, Aaron Nicholls Oculus Perceptual Science Team (Summer 2016)
State of the art eye movement prediction using RNNs, including a new architecture for dealing with noisy data. [figure shows unrelated data from the Roorda Lab] Paper in Progress
Conference Proceedings: A Neural Model of High Acuity Vision in the Presence of Fixational Eye Movements
Alex Anderson, Kavitha Ratnam, Austin Roorda, Bruno Olshausen Asilomar Conference on Signals, Systems, and Computers (Nov 2016)
Model corroborates psychophysical results on benefits of motion for high-acuity vision. Paper
Journal Article: Excess of mutational jackpot events in growing populations due to gene surfing
Diana Fusco, Matti Gralka, Jona Kayser, Alex Anderson, Oskar Hallatschek Nature Communications (Oct 2016)
Mutations propagate differently throughout a population depending on the spatial structure of the habitat. Paper
Technical Report: DeepMovie - Using Optical Flow and Deep Neural Networks to Stylize Movies
Alexander G. Anderson, Cory P. Berg, Daniel P. Mossing, Bruno A. Olshausen Berkeley (May 2016)
Style Transfer for Movies. See the Video for a clip of Star Wars in the Art Style of Vincent Van Gogh Paper . Video
Personal Project: Artistic Style Transfer in TensorFlow
Alexander G. Anderson Berkeley (2016)
Implementation of Artistic Style Transfer in TensorFlow Another Photo
Conference Poster: Modeling High Acuity Vision in the Presence of Fixational Eye Movements
Alexander G. Anderson, Bruno A. Olshausen Computational and Systems Neuroscience (COSYNE) (2016)
Tutorial: Sparse Coding Tutorial
Alexander G. Anderson, Jesse Livezey, Guy Isley, Charles Frye Redwood Center for Theoretical Neuroscience (July, 2015)
Co-wrote code and led a tutorial on Sparse Coding for incoming Berkeley Neuroscience students. Link
Class Project: Analyzing the Internal Dynamics of a LSTM Trained to Solve the Reber Grammar
Alexander G. Anderson Redwood Center for Theoretical Neuroscience (2014)
Implemented a LSTM in theano and analyzed the hidden dynamics that give rise to the solution of the Reber Grammar problem.
Class Project: Parameter Sensitivity for MIRT Models
Alexander G. Anderson, Eliana Feasley (2014)
Final project for a class on graphical modeling. Understood and modified a codebase that used EM (expectation-maximization) to predict student performance on math questions.
Graduate Research Rotation Project: Standing Variation in 2D Range Expansions
Alexander G. Anderson, Matti Gralka, Oskar Hallatschek University of California, Berkeley (2014)
Used a combination of simulations and theory to study the statistical properties of populations that populate a two-dimensional space.
Journal Article: Complex Trajectories in a Classical Periodic Potential
Alexander G Anderson and Carl M Bender Journal of Physics A Mathematical and Theoretical (2012)
Journal Article: Periodic Orbits for Classical Particles Having Complex Energy
Alexander G Anderson, Carl M Bender, and Uriel I. Morone Physics Letters A (2011)
Independent Project: The Algebra of Noncommutative Operators
Alexander G. Anderson, Henry Maltby, Jing Jing Li Berkeley Math Tournament Power Round (2013)
A proof-based round for a high school math contest about an important mathematical tool used in quantum mechanics. Problems | Solutions
Curriculum Design: AwesomeMath Summer Course in Intermediate Geometry
Alexander G. Anderson, Joseph Tkadlec AwesomeMath Summer Camp (2010, 2012)
Fourteen ninty minute lectures and ninty minute problem solving sessions for high school students excited about math competitions.
Summer Research: Phase Determination of Multilayer Mirrors using the Kramers-Kronig Relations
Alexander G. Anderson, Charles Bourassin-Bouchet, Sebastien de Rossi, Franck Delmotte (2011)
Determined that a potential new method of characterizing Multilayer mirrors was infeasible.
Class Project: Isolation and Characterization of Mycobacteriophage Uncle Howie
Alexander G. Anderson Phage Hunters Class (2008)