Table of Contents 7/10ths SynthesisPeter F. Essigpage 5 Aug 1992 (My Small Catechism)Jon Montgome... more Table of Contents 7/10ths SynthesisPeter F. Essigpage 5 Aug 1992 (My Small Catechism)Jon Montgomerypage 6 Chaos Is-J. Dylan McNeillpage 7 UntouchedTraci Williamspage 8 The JustificationJohn C. Carminepage 8 LincolnJon Montgomerypage 9 Untitled (Photo)Nicole Niemanpage 10 Park PoemJohn Brillhartpage 11 SmokeJulia Ann Canhampage 12 Warming the BenchAnn Moutraypage 12 Cereal KillerJay Harnackpage 13 The Dutiful SonsTom McGrathpage 14 UntitledCatherine DeGraafpage 17 7-up bottleWalt Howardpage 17 BreedDan Trutterpage 18 An Argument Against LoveTony Martinezpage 19 UntitledT. Scott Laniganpage 19 Glassblowers BallStephanie Franzenpage 20 Portrait of a Young GirlJohn C. Carminepage 20 Untitled (artwork)Dan Trutterpage 21 Death of a FriendLizabeth Kulkapage 22 Submission BluesMartin Paul Brittpage 23 To the Fourteen Year Old SuicideScott Langenpage 23 The Flabby PilgrimTom McGrathpage 24 The Fall of ImmortalityBrian Wheelerpage 25 Merging with AirThom Schnarrepage 26 UntitledCatherine DeGr...
Chronicles the growth of the cable TV station, from its humble beginnings in a Columbus, Ohio war... more Chronicles the growth of the cable TV station, from its humble beginnings in a Columbus, Ohio warehouse to its current status as a cultural megapower.
TUEXECUTIVE SUMMARY AND PARK PLAN PRIORITIESUT...................................................... more TUEXECUTIVE SUMMARY AND PARK PLAN PRIORITIESUT................................................................ 5
Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task wi... more Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task within a target distribution. A previous theoretical study has argued that this remarkable performance is because the meta-training protocol incentivises agents to behave Bayes-optimally. We empirically investigate this claim on a number of prediction and bandit tasks. Inspired by ideas from theoretical computer science, we show that meta-learned and Bayes-optimal agents not only behave alike, but they even share a similar computational structure, in the sense that one agent system can approximately simulate the other. Furthermore, we show that Bayes-optimal agents are fixed points of the meta-learning dynamics. Our results suggest that memory-based meta-learning might serve as a general technique for numerically approximating Bayes-optimal agents - that is, even for task distributions for which we currently don't possess tractable models.
Genetically encoded calcium indicators have proven useful for characterizing dorsal root ganglion... more Genetically encoded calcium indicators have proven useful for characterizing dorsal root ganglion neuron excitability in vivo. Challenges persist in achieving high spatial-temporal resolutions in vivo, however, due to deep tissue imaging and motion artifacts that may be limiting technical factors in obtaining measurements. Here we report an ex vivo imaging method, using a peripheral neuron-specific Advillin-GCaMP mouse line and electric field stimulation of dorsal root ganglion tissues, to assess the sensitivity of neurons en bloc. The described method rapidly characterizes Ca2+ activity in hundreds of dorsal root ganglion neurons (221 ± 64 per dorsal root ganglion) with minimal perturbation to the in situ soma environment. We further validate the method for use as a drug screening platform with the voltage-gated sodium channel inhibitor, tetrodotoxin. Drug treatment led to decreased evoked Ca2+ activity; half-maximal response voltage (EV50) increased from 13.4 V in untreated tissue...
Probability trees are one of the simplest models of causal generative processes. They possess cle... more Probability trees are one of the simplest models of causal generative processes. They possess clean semantics and -- unlike causal Bayesian networks -- they can represent context-specific causal dependencies, which are necessary for e.g. causal induction. Yet, they have received little attention from the AI and ML community. Here we present concrete algorithms for causal reasoning in discrete probability trees that cover the entire causal hierarchy (association, intervention, and counterfactuals), and operate on arbitrary propositional and causal events. Our work expands the domain of causal reasoning to a very general class of discrete stochastic processes.
Table of Contents 7/10ths SynthesisPeter F. Essigpage 5 Aug 1992 (My Small Catechism)Jon Montgome... more Table of Contents 7/10ths SynthesisPeter F. Essigpage 5 Aug 1992 (My Small Catechism)Jon Montgomerypage 6 Chaos Is-J. Dylan McNeillpage 7 UntouchedTraci Williamspage 8 The JustificationJohn C. Carminepage 8 LincolnJon Montgomerypage 9 Untitled (Photo)Nicole Niemanpage 10 Park PoemJohn Brillhartpage 11 SmokeJulia Ann Canhampage 12 Warming the BenchAnn Moutraypage 12 Cereal KillerJay Harnackpage 13 The Dutiful SonsTom McGrathpage 14 UntitledCatherine DeGraafpage 17 7-up bottleWalt Howardpage 17 BreedDan Trutterpage 18 An Argument Against LoveTony Martinezpage 19 UntitledT. Scott Laniganpage 19 Glassblowers BallStephanie Franzenpage 20 Portrait of a Young GirlJohn C. Carminepage 20 Untitled (artwork)Dan Trutterpage 21 Death of a FriendLizabeth Kulkapage 22 Submission BluesMartin Paul Brittpage 23 To the Fourteen Year Old SuicideScott Langenpage 23 The Flabby PilgrimTom McGrathpage 24 The Fall of ImmortalityBrian Wheelerpage 25 Merging with AirThom Schnarrepage 26 UntitledCatherine DeGr...
Chronicles the growth of the cable TV station, from its humble beginnings in a Columbus, Ohio war... more Chronicles the growth of the cable TV station, from its humble beginnings in a Columbus, Ohio warehouse to its current status as a cultural megapower.
TUEXECUTIVE SUMMARY AND PARK PLAN PRIORITIESUT...................................................... more TUEXECUTIVE SUMMARY AND PARK PLAN PRIORITIESUT................................................................ 5
Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task wi... more Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task within a target distribution. A previous theoretical study has argued that this remarkable performance is because the meta-training protocol incentivises agents to behave Bayes-optimally. We empirically investigate this claim on a number of prediction and bandit tasks. Inspired by ideas from theoretical computer science, we show that meta-learned and Bayes-optimal agents not only behave alike, but they even share a similar computational structure, in the sense that one agent system can approximately simulate the other. Furthermore, we show that Bayes-optimal agents are fixed points of the meta-learning dynamics. Our results suggest that memory-based meta-learning might serve as a general technique for numerically approximating Bayes-optimal agents - that is, even for task distributions for which we currently don't possess tractable models.
Genetically encoded calcium indicators have proven useful for characterizing dorsal root ganglion... more Genetically encoded calcium indicators have proven useful for characterizing dorsal root ganglion neuron excitability in vivo. Challenges persist in achieving high spatial-temporal resolutions in vivo, however, due to deep tissue imaging and motion artifacts that may be limiting technical factors in obtaining measurements. Here we report an ex vivo imaging method, using a peripheral neuron-specific Advillin-GCaMP mouse line and electric field stimulation of dorsal root ganglion tissues, to assess the sensitivity of neurons en bloc. The described method rapidly characterizes Ca2+ activity in hundreds of dorsal root ganglion neurons (221 ± 64 per dorsal root ganglion) with minimal perturbation to the in situ soma environment. We further validate the method for use as a drug screening platform with the voltage-gated sodium channel inhibitor, tetrodotoxin. Drug treatment led to decreased evoked Ca2+ activity; half-maximal response voltage (EV50) increased from 13.4 V in untreated tissue...
Probability trees are one of the simplest models of causal generative processes. They possess cle... more Probability trees are one of the simplest models of causal generative processes. They possess clean semantics and -- unlike causal Bayesian networks -- they can represent context-specific causal dependencies, which are necessary for e.g. causal induction. Yet, they have received little attention from the AI and ML community. Here we present concrete algorithms for causal reasoning in discrete probability trees that cover the entire causal hierarchy (association, intervention, and counterfactuals), and operate on arbitrary propositional and causal events. Our work expands the domain of causal reasoning to a very general class of discrete stochastic processes.
Uploads
Papers by Tom McGrath