### Omnidirectional Lifelong Learning via Ensembling Representations [JHU](https://www.jhu.edu/): Jayanta Dey | Ali Geisa | Hayden Helm | Ronak Mehta | Will LeVine | Carey E. Priebe | Joshua T. Vogelstein
[Microsoft Research](https://www.microsoft.com/en-us/research/): Weiwei Yang | Jonathan Larson | Bryan Tower | Chris White ![:scale 40%](images/neurodata_blue.png) --- ### Different Learning Paradigms - biological learning: continually use new data to improve performance on both .ye[past] and potential .ye[future] tasks - (classical) machine learning: .ye[tabula rasa], forget old tasks with new data --- ### Existing strategies - Given new data, either - forward transfer: improve performance on .ye[future tasks] given current data - avoid catastrophic forgetting: do not .ye[forget past] tasks - These goals are .ye[inadequate] - We desire .ye[synergistic learning]: transfer both forward and backward --- ### Our approach: Ensembling Representations ![:scale 100%](images/learning_schema_new.png) --- ### Omnidirectional Algorithms can Transfer Between XOR and XNOR ![:scale 100%](images/xor_xnor_exp.png) --- ### CIFAR 10x10 .pull-left[ - *CIFAR 100* is a popular image classification dataset with 100 classes of images. - 500 training images and 100 testing images per class. - All images are 32x32 color images. - CIFAR 10x10 breaks the 100-class task problem into 10 tasks, each with 10-class. ] .pull-right[
] --- ### Omnidirectional Algorithms Show Forward Transfer for the CIFAR 10x10 Tasks ![:scale 115%](images/cifar_exp_fte.png) --- ### Omnidirectional Algorithms Uniquely Show Backward Transfer for Each CIFAR 10x10 Task ![:scale 115%](images/cifar_exp_bte.png) --- ### Acknowledgements
yummy
lion
baby girl
family
earth
milkyway
##### JHU
Carey Priebe
Jesse Patsolic
Meghana Madhya
Hayden Helm
Richard Gou
Ronak Mehta
Jayanta Dey
Will LeVine
##### Microsoft Research
Chris White
Weiwei Yang
Jonathan Larson
Bryan Tower
##### DARPA L2M {[BME](https://www.bme.jhu.edu/),[CIS](http://cis.jhu.edu/), [ICM](https://icm.jhu.edu/), [KNDI](http://kavlijhu.org/)}@[JHU](https://www.jhu.edu/) | [neurodata](https://neurodata.io)
[jovo@jhu.edu](mailto:j1c@jhu.edu) |
| [@neuro_data](https://twitter.com/neuro_data) --- background-image: url(images/l_and_v.jpeg) .footnote[Questions?]