-
Towards Efficient and Scalable Training of Differentially Private Deep Learning
Authors:
Sebastian Rodriguez Beltran,
Marlon Tobaben,
Niki Loppi,
Antti Honkela
Abstract:
Differentially private stochastic gradient descent (DP-SGD) is the standard algorithm for training machine learning models under differential privacy (DP). The major drawback of DP-SGD is the drop in utility which prior work has comprehensively studied. However, in practice another major drawback that hinders the large-scale deployment is the significantly higher computational cost. We conduct a c…
▽ More
Differentially private stochastic gradient descent (DP-SGD) is the standard algorithm for training machine learning models under differential privacy (DP). The major drawback of DP-SGD is the drop in utility which prior work has comprehensively studied. However, in practice another major drawback that hinders the large-scale deployment is the significantly higher computational cost. We conduct a comprehensive empirical study to quantify the computational cost of training deep learning models under DP and benchmark methods that aim at reducing the cost. Among these are more efficient implementations of DP-SGD and training with lower precision. Finally, we study the scaling behaviour using up to 80 GPUs.
△ Less
Submitted 25 June, 2024;
originally announced June 2024.
-
Expansion of Visual Hints for Improved Generalization in Stereo Matching
Authors:
Andrea Pilzer,
Yuxin Hou,
Niki Loppi,
Arno Solin,
Juho Kannala
Abstract:
We introduce visual hints expansion for guiding stereo matching to improve generalization. Our work is motivated by the robustness of Visual Inertial Odometry (VIO) in computer vision and robotics, where a sparse and unevenly distributed set of feature points characterizes a scene. To improve stereo matching, we propose to elevate 2D hints to 3D points. These sparse and unevenly distributed 3D vis…
▽ More
We introduce visual hints expansion for guiding stereo matching to improve generalization. Our work is motivated by the robustness of Visual Inertial Odometry (VIO) in computer vision and robotics, where a sparse and unevenly distributed set of feature points characterizes a scene. To improve stereo matching, we propose to elevate 2D hints to 3D points. These sparse and unevenly distributed 3D visual hints are expanded using a 3D random geometric graph, which enhances the learning and inference process. We evaluate our proposal on multiple widely adopted benchmarks and show improved performance without access to additional sensors other than the image sequence. To highlight practical applicability and symbiosis with visual odometry, we demonstrate how our methods run on embedded hardware.
△ Less
Submitted 1 November, 2022;
originally announced November 2022.
-
Spatio-Temporal Variational Gaussian Processes
Authors:
Oliver Hamelijnck,
William J. Wilkinson,
Niki A. Loppi,
Arno Solin,
Theodoros Damoulas
Abstract:
We introduce a scalable approach to Gaussian process inference that combines spatio-temporal filtering with natural gradient variational inference, resulting in a non-conjugate GP method for multivariate data that scales linearly with respect to time. Our natural gradient approach enables application of parallel filtering and smoothing, further reducing the temporal span complexity to be logarithm…
▽ More
We introduce a scalable approach to Gaussian process inference that combines spatio-temporal filtering with natural gradient variational inference, resulting in a non-conjugate GP method for multivariate data that scales linearly with respect to time. Our natural gradient approach enables application of parallel filtering and smoothing, further reducing the temporal span complexity to be logarithmic in the number of time steps. We derive a sparse approximation that constructs a state-space model over a reduced set of spatial inducing points, and show that for separable Markov kernels the full and sparse cases exactly recover the standard variational GP, whilst exhibiting favourable computational properties. To further improve the spatial scaling we propose a mean-field assumption of independence between spatial locations which, when coupled with sparsity and parallelisation, leads to an efficient and accurate method for large spatio-temporal problems.
△ Less
Submitted 2 November, 2021;
originally announced November 2021.
-
D3p -- A Python Package for Differentially-Private Probabilistic Programming
Authors:
Lukas Prediger,
Niki Loppi,
Samuel Kaski,
Antti Honkela
Abstract:
We present d3p, a software package designed to help fielding runtime efficient widely-applicable Bayesian inference under differential privacy guarantees. d3p achieves general applicability to a wide range of probabilistic modelling problems by implementing the differentially private variational inference algorithm, allowing users to fit any parametric probabilistic model with a differentiable den…
▽ More
We present d3p, a software package designed to help fielding runtime efficient widely-applicable Bayesian inference under differential privacy guarantees. d3p achieves general applicability to a wide range of probabilistic modelling problems by implementing the differentially private variational inference algorithm, allowing users to fit any parametric probabilistic model with a differentiable density function. d3p adopts the probabilistic programming paradigm as a powerful way for the user to flexibly define such models. We demonstrate the use of our software on a hierarchical logistic regression example, showing the expressiveness of the modelling approach as well as the ease of running the parameter inference. We also perform an empirical evaluation of the runtime of the private inference on a complex model and find a $\sim$10 fold speed-up compared to an implementation using TensorFlow Privacy.
△ Less
Submitted 15 September, 2021; v1 submitted 22 March, 2021;
originally announced March 2021.
