LIVIVO - The Search Portal for Life Sciences

zur deutschen Oberfläche wechseln
Advanced search

Search results

Result 1 - 3 of total 3

Search options

  1. Book ; Online: ZipIt! Merging Models from Different Tasks without Training

    Stoica, George / Bolya, Daniel / Bjorner, Jakob / Hearn, Taylor / Hoffman, Judy

    2023  

    Abstract: Typical deep visual recognition models are capable of performing the one task they were trained on. In this paper, we tackle the extremely difficult problem of combining completely distinct models with different initializations, each solving a separate ... ...

    Abstract Typical deep visual recognition models are capable of performing the one task they were trained on. In this paper, we tackle the extremely difficult problem of combining completely distinct models with different initializations, each solving a separate task, into one multi-task model without any additional training. Prior work in model merging permutes one model to the space of the other then adds them together. While this works for models trained on the same task, we find that this fails to account for the differences in models trained on disjoint tasks. Thus, we introduce "ZipIt!", a general method for merging two arbitrary models of the same architecture that incorporates two simple strategies. First, in order to account for features that aren't shared between models, we expand the model merging problem to additionally allow for merging features within each model by defining a general "zip" operation. Second, we add support for partially zipping the models up until a specified layer, naturally creating a multi-head model. We find that these two changes combined account for a staggering 20-60% improvement over prior work, making the merging of models trained on disjoint tasks feasible.
    Keywords Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Machine Learning
    Subject code 004
    Publishing date 2023-05-04
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  2. Article ; Online: Separate populations of neurons in ventral striatum encode value and motivation.

    Bissonette, Gregory B / Burton, Amanda C / Gentry, Ronny N / Goldstein, Brandon L / Hearn, Taylor N / Barnett, Brian R / Kashtelyan, Vadim / Roesch, Matthew R

    PloS one

    2013  Volume 8, Issue 5, Page(s) e64673

    Abstract: Neurons in the ventral striatum (VS) fire to cues that predict differently valued rewards. It is unclear whether this activity represents the value associated with the expected reward or the level of motivation induced by reward anticipation. To ... ...

    Abstract Neurons in the ventral striatum (VS) fire to cues that predict differently valued rewards. It is unclear whether this activity represents the value associated with the expected reward or the level of motivation induced by reward anticipation. To distinguish between the two, we trained rats on a task in which we varied value independently from motivation by manipulating the size of the reward expected on correct trials and the threat of punishment expected upon errors. We found that separate populations of neurons in VS encode expected value and motivation.
    MeSH term(s) Action Potentials/physiology ; Animals ; Basal Ganglia/cytology ; Behavior, Animal/physiology ; Male ; Motivation ; Motor Activity/physiology ; Neurons/cytology ; Neurons/physiology ; Rats ; Rats, Long-Evans ; Regression Analysis ; Reward ; Task Performance and Analysis
    Language English
    Publishing date 2013-05-28
    Publishing country United States
    Document type Journal Article ; Research Support, N.I.H., Extramural
    ISSN 1932-6203
    ISSN (online) 1932-6203
    DOI 10.1371/journal.pone.0064673
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

  3. Article ; Online: Prenatal Nicotine Exposure Impairs Executive Control Signals in Medial Prefrontal Cortex.

    Bryden, Daniel W / Burton, Amanda C / Barnett, Brian R / Cohen, Valerie J / Hearn, Taylor N / Jones, Emily A / Kariyil, Reshma J / Kunin, Alice / Kwak, Sae In / Lee, Jessica / Lubinski, Brooke L / Rao, Gautam K / Zhan, Ashley / Roesch, Matthew R

    Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology

    2016  Volume 41, Issue 3, Page(s) 716–725

    Abstract: Prenatal nicotine exposure (PNE) is linked to numerous psychiatric disorders including attention deficit hyperactivity disorder (ADHD). Current literature suggests that core deficits observed in ADHD reflect abnormal inhibitory control governed by the ... ...

    Abstract Prenatal nicotine exposure (PNE) is linked to numerous psychiatric disorders including attention deficit hyperactivity disorder (ADHD). Current literature suggests that core deficits observed in ADHD reflect abnormal inhibitory control governed by the prefrontal cortex. Yet, it is unclear how neural activity in the medial prefrontal cortex (mPFC) is modulated during tasks that assess response inhibition or if these neural correlates, along with behavior, are affected by PNE. To address this issue, we recorded from single mPFC neurons in control and PNE rats as they performed a stop-signal task. We found that PNE rats were faster for all trial-types, made more premature responses, and were less likely to inhibit behavior on 'STOP' trials during which rats had to inhibit an already initiated response. Activity in mPFC was modulated by response direction and was positively correlated with accuracy and movement time in control but not PNE rats. Although the number of single neurons correlated with response direction was significantly reduced by PNE, neural activity observed on general STOP trials was largely unaffected. However, dramatic behavioral deficits on STOP trials immediately following non-conflicting (GO) trials in the PNE group appear to be mediated by the loss of conflict monitoring signals in mPFC. We conclude that prenatal nicotine exposure makes rats impulsive and disrupts firing of mPFC neurons that carry signals related to response direction and conflict monitoring.
    MeSH term(s) Action Potentials ; Animals ; Electrodes, Implanted ; Executive Function/drug effects ; Executive Function/physiology ; Female ; Inhibition (Psychology) ; Male ; Motor Activity/drug effects ; Motor Activity/physiology ; Neurons/drug effects ; Neurons/physiology ; Neuropsychological Tests ; Nicotine/toxicity ; Nicotinic Agonists/toxicity ; Prefrontal Cortex/drug effects ; Prefrontal Cortex/growth & development ; Prefrontal Cortex/physiopathology ; Pregnancy ; Prenatal Exposure Delayed Effects ; Rats, Long-Evans
    Chemical Substances Nicotinic Agonists ; Nicotine (6M3C89ZY6R)
    Language English
    Publishing date 2016-02
    Publishing country England
    Document type Journal Article ; Research Support, N.I.H., Extramural ; Research Support, Non-U.S. Gov't
    ZDB-ID 639471-1
    ISSN 1740-634X ; 0893-133X
    ISSN (online) 1740-634X
    ISSN 0893-133X
    DOI 10.1038/npp.2015.197
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

To top