diff --git a/_speakers/5_aldopacchiano.md b/_speakers/5_aldopacchiano.md new file mode 100644 index 0000000..41fc8da --- /dev/null +++ b/_speakers/5_aldopacchiano.md @@ -0,0 +1,24 @@ +--- +# Name of the speaker +name: Aldo Pacchiano + +# Link to the speaker's webpage +webpage: https://www.aldopacchiano.ai/ + +# Primary affiliation of the speaker +affil: Broad Institute +# Link to the speaker's primary affiliation +affil_link: https://www.broadinstitute.org/ + +# An image of the speaker (square aspect ratio works the best) (place in the `assets/img/speakers` directory) +img: aldopacchiano.jpeg + +# # (Optional) Secondary affiliation of the speaker +# affil2: BuzzFizz Corp +# # Link to the speaker's secondary affiliation +# affil2_link: https://buzzfizz.corp +--- + + + +In settings where there is a significant overhead to deploying adaptive algorithms -- for example, when the execution of the data collection policies is required to be distributed, or a human in the loop is needed to implement these policies -- producing in advance a set of policies for data collection is paramount. We study the setting where a large dataset of contexts but not rewards is available and may be used by the learner to design an effective data collection strategy. Although when rewards are linear this problem has been well studied, results are still missing for more complex reward models. In this work we propose two experiment planning strategies compatible with function approximation. The first is an eluder planning and sampling procedure that can recover optimality guarantees depending on the eluder dimension of the reward function class. For the second, we show that a uniform sampler achieves competitive optimality rates in the setting where the number of actions is small. We finalize our results introducing a statistical gap fleshing out the fundamental differences between planning and adaptive learning and provide results for planning with model selection. diff --git a/_speakers/6_katedarling.md b/_speakers/6_katedarling.md new file mode 100644 index 0000000..2783172 --- /dev/null +++ b/_speakers/6_katedarling.md @@ -0,0 +1,24 @@ +--- +# Name of the speaker +name: Kate Darling + +# Link to the speaker's webpage +webpage: https://www.katedarling.org/ + +# Primary affiliation of the speaker +affil: MIT Media Lab +# Link to the speaker's primary affiliation +affil_link: https://www.media.mit.edu/ + +# An image of the speaker (square aspect ratio works the best) (place in the `assets/img/speakers` directory) +img: katedarling.jpeg + +# # (Optional) Secondary affiliation of the speaker +# affil2: BuzzFizz Corp +# # Link to the speaker's secondary affiliation +# affil2_link: https://buzzfizz.corp +--- + + + +__TBD__ diff --git a/assets/abstracts/aldopacchiano.txt b/assets/abstracts/aldopacchiano.txt new file mode 100644 index 0000000..024dc73 --- /dev/null +++ b/assets/abstracts/aldopacchiano.txt @@ -0,0 +1 @@ +In settings where there is a significant overhead to deploying adaptive algorithms -- for example, when the execution of the data collection policies is required to be distributed, or a human in the loop is needed to implement these policies -- producing in advance a set of policies for data collection is paramount. We study the setting where a large dataset of contexts but not rewards is available and may be used by the learner to design an effective data collection strategy. Although when rewards are linear this problem has been well studied, results are still missing for more complex reward models. In this work we propose two experiment planning strategies compatible with function approximation. The first is an eluder planning and sampling procedure that can recover optimality guarantees depending on the eluder dimension of the reward function class. For the second, we show that a uniform sampler achieves competitive optimality rates in the setting where the number of actions is small. We finalize our results introducing a statistical gap fleshing out the fundamental differences between planning and adaptive learning and provide results for planning with model selection. \ No newline at end of file diff --git a/assets/img/speakers/aldopacchiano.jpeg b/assets/img/speakers/aldopacchiano.jpeg new file mode 100644 index 0000000..1ad6489 Binary files /dev/null and b/assets/img/speakers/aldopacchiano.jpeg differ diff --git a/assets/img/speakers/katedarling.jpeg b/assets/img/speakers/katedarling.jpeg new file mode 100644 index 0000000..2c9b0b7 Binary files /dev/null and b/assets/img/speakers/katedarling.jpeg differ diff --git a/index.md b/index.md index 8ec64e5..ea93ee5 100644 --- a/index.md +++ b/index.md @@ -46,5 +46,15 @@ Brown Robotics Talks consists of BigAI talks and lab talks ([CIT](https://www.go TBD Jiayuan Mao + + 04/26 + Experiment Planning with Function Approximation [abstract] + Aldo Pacchiano + + + 05/10 + TBD + Kate Darling + \ No newline at end of file