Lawson L.S. Wong of MIT will present “Learning The State Of The World Object-based World Modeling For Mobile-Manipulation Robots” Thursday Dec. 3rd from 1:00pm to 2:00pm (Arrive at 12:45 to get pizza) in CS150. An abstract and bio follow.

Abstract:

Mobile-manipulation robots performing service tasks in human-centric indoor environments need to know about relevant aspects of their spatial surroundings. However, service robots rarely know the exact state of the world, unlike industrial robots in structured environments. Additionally, as the world is shared with humans, uncertainty in the complete state of the world is inevitable over time. Mobile-manipulation robots therefore need to continuously perform state estimation, using perceptual information to maintain a representation of the state, and its uncertainty, of task-relevant aspects of the world. Because indoor tasks frequently require interacting with objects, objects should be given critical emphasis in spatial representations for service robots. In my Ph.D. work, I propose a world model based on objects, their semantic attributes (task-relevant properties such as type and pose), and their geometric realizations in the physical world.

Objects are challenging to keep track of because there is significant uncertainty in their states. Object detection and recognition using robotic vision is still error-prone. Objects can also be inherently ambiguous because they have similar attributes. Besides detection noise, other agents may change the state of the world. Compounded over multitudes of objects and long temporal horizons, the above sources of uncertainty give rise to a challenging estimation problem. Fortunately, most objects do not change quickly, and sensing is relatively cheap, so we can leverage information from multiple diverse snapshots of similar world states. However, putting the information together introduces a data association problem, which I tackle with constrained Bayesian nonparametric models. By carefully aggregating information across different viewpoints, times, and sensors, I show that robots can reduce their uncertainty in the state of the world and maintain more accurate object-based world models.

Bio:

Lawson L.S. Wong is a Ph.D. candidate at the Massachusetts Institute of Technology, working in the Learning and Intelligent Systems Group under the supervision of Leslie Pack Kaelbling and Tomás Lozano-Pérez. Previously, he received his B.S. (with Honors) and M.S. in Computer Science at Stanford University, both in 2009. His current research focuses on acquiring, representing, and estimating knowledge about the world that an autonomous robot may find useful. More broadly, Lawson is interested in, and follows many topics within, the fields of robotics, machine learning, and artificial intelligence. He was recently awarded a AAAI Robotics Student Fellowship and a Croucher Foundation Fellowship for Postdoctoral Research. He will begin his postdoctoral appointment at Brown University in 2016, working with Stefanie Tellex.

Jon Sprouse (UConn) will give the Linguistics department colloquium at 3:30 on Friday, December 4, in ILC 400. His research focuses on questions about syntactic representations and operations, with a particular focus on experimental approaches to investigating these questions.

Louise Antony of UMass Philosophy will present “In Praise of Loose Talk: on the notion of ‘rule-following’ in cognitive science” in the Cognitive Brown Bag series Wednesday Dec. 12th in Tobin 521B. The abstract follows.

Abstract: Philosophers often challenge cognitive scientists to clarify their foundational assumptions, and in particular, to clarify their use of intentionalistic terms like “knowledge,” “representation,” and “inference.”  Cognitive scientists often disparage these challenges, saying that any needed clarifications will come as empirical work progresses.  This is a conciliatory paper.  On the one hand I’ll argue, through a case study from vision science, that empirical inquiry need not wait for the clarification of its foundations, and that empirical scientists should not be pressed to define terms that are serving their needs perfectly well.  On the other hand, I’ll show that some philosophical clarification is both possible and salutary.  I’ll offer a taxonomy of “rule-following” cognitive systems – “rational-causal,” “intelligible-causal,” and “brute causal” systems – which shows how differences in cognitive architecture might align with distinctions philosophers deem important.

David Blei of Columbia University will be giving a talk in Computer Science 150/151 on Thursday 11/19 from 5 to 6 pm on “Scaling and Generalizing Variational Inference”. He will also give a talk to the Computational Social Science Initiative on Friday 11/20 from 12:30 – 2 in Lederle 1634 on  “Probabilistic Topic Models and User Behavior”.

Marcus Maia and Aniela Franca of the Federal University of Rio de Janeiro will present in the Language Acquisition Research Center this Friday 11/20 at noon in ILC N451. All are welcome! Two talks will be presented:

A Computational Efficiency Principle in action in the processing of recursively embedded PPs in Brazilian Portuguese and in Karaja  – Marcus Maia (Federal University of Rio de Janeiro)

An ERP Evaluation of the comprehension of PP embedding and coordination in Karaja and in Portuguese – Aniela França (Federal University of Rio de Janeiro)

A workshop on Turkish linguistics is being held in the Department of Linguistics Saturday and Sunday November 21st – 22nd. All are welcome, and a full program is here.

Laura McPherson (Dartmouth College) will give a Linguistics colloquium this Friday at 3:30 in ILC N400. The title of her talk is “Constraint Interaction and the role of spell-out in Dogon tonosyntax,” and an abstract follows.

The Dogon languages of Mali share a unique system of replacive grammatical tone in the DP, where a word’s lexical tone is completely overwritten by tonal overlays in particular morphosyntactic positions. Unlike more typologically common systems of replacive tone, which tend to be triggered by morphemes or morphological features and are confined to a single word, Dogon overlays in the DP may span multiple words and are triggered by certain c-commanding syntactic categories or positions (hence, tonosyntax). In cases where a word is targeted by more than one potential trigger, the Dogon languages differ in their resolutions. I argue that these changes are inherently morphological in nature, despite occurring at the phrase-level rather than the word-level, and propose a construction-based model in which phrase-level morphological constructions encode idiosyncratic phonological changes sensitive to both syntactic category and syntactic structure. Constructional schemas are implemented in the grammar as constraints (construction constraints). The variation found in the Dogon language family is descriptively quite complicated, with no two languages working in exactly the same way, but in this talk, I show that the surface patterns falls out naturally in a maximum entropy (Goldwater and Johnson 2003, Hayes and Wilson 2008) model with weighted constraints, capable of capturing both within-language and between-language variation.

Additionally, data from the tonosyntax of possession and relative clauses provide evidence for the role of phases in determining morphophonological form. In particular, the application of tonal overlays is often, though (crucially) not obligatorily, blocked on material that has spelled out in a previous cycle. I argue that these effects provide evidence for transcyclic faithfulness constraints, which penalize alterations to the morphophonological form of spelled out material in later grammatical cycles. Like all constraints, these too are shown to be violable, with different Dogon languages displaying varying degrees of faithfulness to phasal targets. Thus, the Dogon data show that while spelled out material may be resistant to later phonological changes, it is not immune to it, as argued by proposals such as Lowenstamm (2010) or Newell and Piggott (2014).

As announced last week, a group of us will meet this Thursday to discuss Boekel et al.’s (2015) Cortex paper that attempts to replicate a set of structural brain-behavior correlations. All are welcome to join us. The meeting will take place at 12:45 in ILC N400. Contrary to the prior announcement, we will not go to the dining commons, but instead will stay at N400 and nosh on People’s Market bagels and cream cheese. Coffee and tea will also be provided. It’d still be great if those of you planning to participate could please fill in your name and department in the google form linked here, though others are also welcome, and we will have a little extra.

Wouter Boekel, Eric-Jan Wagenmakers, Luam Belay, Josine Verhagen, Scott Brown and Birte U. Forstmann. 2015. A purely confirmatory replication study of structural brain-behavior correlations. Cortex.

In a little over a week on Thursday Nov. 19th at 12:45, we will have a discussion of the paper below, linked here, over lunch. We’ll meet in Linguistics, outside of ILC N400, and head over to the Worcester Dining Common. A lunch ticket will be provided for all participants. If you wish to participate, please fill in your name and department in the google form linked here before 9 am 11/19 (we’ll get tickets that morning). This will likely be the first of a series of these meetings, which we’ve initiated based on an idea that came out of a graduate student meeting. Assuming that there is interest, next semester we will poll more broadly to find a time. Please contact Joe Pater for more information.

Wouter Boekel, Eric-Jan Wagenmakers, Luam Belay, Josine Verhagen, Scott Brown and Birte U. Forstmann. 2015. A purely confirmatory replication study of structural brain-behavior correlations. Cortex.

Howard Lasnik (University of Maryland) will give a colloquium on Friday at 3:30 in ILC N400. The title of his talk is “Clause-mates, Phases and Two Families of Questions.” An abstract is here.