Safe mutations for deep and recurrent neural networks through output gradients
Open accessThe DOI matches its Crossref record; 11 of the 11 references it cites were found in a registry or an index.
Abstract
While neuroevolution (evolving neural networks) has been successful across a variety of domains from reinforcement learning, to artificial life, to evolutionary robotics, it is rarely applied to large, deep neural networks. A central reason is that while random mutation generally works in low dimensions, a random perturbation of thousands or millions of weights will likely break existing functionality. This paper proposes a solution: a family of safe mutation (SM) operators that facilitate exploration without dramatically altering network behavior or requiring additional interaction with the environment. The most effective SM variant scales the degree of mutation of each individual weight according to the sensitivity of the network's outputs to that weight, which requires computing the gradient of outputs with respect to the weights (instead of the gradient of error, as in conventional deep learning). This safe mutation through gradients (SM-G) operator dramatically increases the ability of a simple genetic algorithm-based neuroevolution method to find solutions in high-dimensional domains that require deep and/or recurrent neural networks, including domains that require processing raw pixels. By improving our ability to evolve deep neural networks, this new safer approach to mutation expands the scope of domains amenable to neuroevolution.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Zagoruyko, S., & Komodakis, N. (2016) · Procedings of the British Machine Vision Conference 2016 · doi:10.5244/c.30.87 · Cited by 7,244
A direct adaptive method for faster backpropagation learning: the RPROP algorithm
Riedmiller, M., & Braun, H. (2002) · IEEE International Conference on Neural Networks · doi:10.1109/icnn.1993.298623 · Cited by 3,917 · cited 2×
“Interestingly, both approaches relate to mechanisms found effective in deep learning, such as adaptive-learning rate methods (Kingma and Ba, 2014; Riedmiller and Braun, 1993) or trust regions in policy gradients (Schulman et al., 2015), although here they manifest through different motivation in a…”
BackgroundHansen, N., Müller, S. D., & Koumoutsakos, P. (2003) · Evolutionary Computation · doi:10.1162/106365603321828970 · Cited by 2,601 · cited 1×
“Related to NES and EDAs, CMA-ES learns to model pair-wise dependencies among parameters, aimed at generating productive offspring (Hansen et al., 2003).”
BackgroundSelf-Normalizing Neural Networks
Klambauer, G., Unterthiner, T., Mayr, A., & Hochreiter, S. (2017) · arXiv · doi:10.48550/arxiv.1706.02515 · Cited by 2,267
Curiosity-Driven Exploration by Self-Supervised Prediction
Pathak, D., Agrawal, P., Efros, A. A., & Darrell, T. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2017.70 · Cited by 1,815 · cited 1×
“…that enables learning abstract representations without feature engineering as well as integrating information over time; similar combinations of recurrence and convolution have shown considerable promise in deep RL (Hausknecht and Stone, 2015; Pathak et al., 2017; Mirowski et al., 2016).”
BackgroundNeuroevolution: from architectures to learning
Floreano, D., Dürr, P., & Mattiussi, C. (2008) · Evolutionary Intelligence · doi:10.1007/s12065-007-0002-4 · Cited by 734 · cited 2×
“r ability to evolve deep neural networks, this new safer approach to mutation expands the scope of domains amenable to neuroevolution. 1 Introduction Neuroevolution (NE; Lehman and Miikkulainen 2013; Floreano et al. 2008; Yao 1999) combines evolutionary algorithms (EAs) and artificial neural networks. It has long been…”
BackgroundResilient Machines Through Continuous Self-Modeling
Bongard, J., Zykov, V., & Lipson, H. (2006) · Science · doi:10.1126/science.1133687 · Cited by 709 · cited 2×
“It has long been popular when gradient information with respect to the task is unavailable or difficult to obtain, e.g. in artificial life (Bedau et al., 2000; Langton, 1989) and evolutionary robotics (Nolfi and Floreano, 2000; Bongard et al., 2006; Cully et al., 2015).”
BackgroundQuality Diversity: A New Frontier for Evolutionary Computation
Pugh, J. K., Soros, L. B., & Stanley, K. O. (2016) · Frontiers in Robotics and AI · doi:10.3389/frobt.2016.00040 · Cited by 514 · cited 2×
“techniques also make no assumptions about the stationarity or uni-modality of reward, and can thus be easily applied to divergent search techniques (Lehman and Stanley, 2011a; Mouret and Clune, 2015; Pugh et al., 2016) and coevolution (Popovici et al., 2012). Additionally, SM-G may productively interact with different…”
BackgroundSelf-citation
Cited by
Activation Functions: Comparison of trends in Practice and Research for Deep Learning
Nwankpa, C. E., Ijomah, W., Gachagan, A., & Marshall, S. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.03378 · Cited by 1,240
Designing neural networks through neuroevolution
Stanley, K. O., Clune, J., Lehman, J., & Miikkulainen, R. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-018-0006-z · Cited by 689
Self-citationAntonopoulos, I., Robu, V., Couraud, B., Kirli, D., Norbu, S., Kiprakis, A., Flynn, D., Elizondo-Gonzalez, S., & Wattam, S. (2020) · Renewable and Sustainable Energy Reviews · doi:10.1016/j.rser.2020.109899 · Cited by 656
Artificial Intelligence Enabled Demand Response: Prospects and Challenges in Smart Grid Environment
Khan, M. A., Saleh, A. M., Waseem, M., & Sajjad, I. A. (2023) · IEEE Access · doi:10.1109/access.2022.3231444 · Cited by 179
DeepMasterPrints: Generating MasterPrints for Dictionary Attacks via Latent Variable Evolution
Bontrager, P., Roy, A., Togelius, J., Memon, N., & Ross, A. (2018) · 2018 IEEE 9th International Conference on Biometrics Theory, Applications and Systems (BTAS) · doi:10.1109/btas.2018.8698539 · Cited by 161
Metadata
- Authors
- Lehman, J., Chen, J., Clune, J., & Stanley, K. O.
- Venue
- Proceedings of the Genetic and Evolutionary Computation Conference
- Publisher
- ACM
- Published
- 2018-07-02
- Pages
- 117-124
- OpenAlex
- W2779918130
- Cited by
- 78
- References
- 11
- FWCI
- 4.47
- Type
- conference paper
- Language
- en
- Open access
- bronze
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.