Guide-yakabatana mibvunzo · Zvakaoma Nhanho

Superposition uye Polysemancity Quiz

Macheki anoratidza superposition, sparsity, polysemantic neurons, anopindirana mafambiro uye geometry yekumisikidzwa inomiririra.

Nzira dzekutungamira dzinoenderanaSuperposition uye Polysemancity
Mubvunzo 1 ye 8

Chii chinonzi 'superposition' chinoreva mumaneural network?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Chii chinonzi 'superposition' chinoreva mumaneural network?
    • Inomiririra mamwe maficha kupfuura ane neuroni nekupindirana kwadzo
    • Kurongedza akawanda mamodheru pamusoro peumwe neumwe
    • Kumhanyisa modhi kaviri uye neavhareji
    • Kuwedzera ruzha kune zvinopinda panguva yekudzidziswa
  2. Chii chinoita kuti superposition ishande zvakanaka kunyangwe kukanganiswa kwechinhu?
    • Zvimiro zviri kushanda zvakanyanya nguva dzese
    • Zvimiro zviri zvishoma, saka vashoma vari kushanda panguva imwe chete
    • Kushandisa ma activations akanaka chete
    • Kuva nema neurons akawanda kupfuura maficha
  3. Chii chinonzi 'polysemantic' neuron?
    • Neuron isingamboshande
    • Neuron inoshandiswa chete muchikamu chekubuda
    • Neuron inopindura kune dzakawanda pfungwa dzisina hukama
    • Neuron isina kubatana
  4. Nderipi Anthropic bepa rakaunza ongororo yakadzorwa yechiitiko ichi muna 2022?
    • Kuchenjerera Ndiko Kwese Kwaunoda
    • ImageNet Classification ine Deep CNNs
    • Kutamba Atari neDeep RL
    • Toy Models yeSuperposition
  5. Kana dhizaini ichifanira kuchengetedza mamwe maficha kupfuura ane hukuru, chii chisingadzivisike chejometri?
    • Mamwe maficha anofanira kugovera anopindirana, asiri-orthogonal nzira
    • Iwo maficha anogona kugara akakwana orthogonal
    • Network inofanira kuparara
    • Mamiriro ese anoita zvakafanana
  6. Sei superposition ichiita kuti kududzira kuome?
    • Inoita kuti modhi inonoke kumhanya
    • Single neurons hachisiri mepu yakachena kune imwechete pfungwa
    • Inobvisa zvinhu zvose kubva pamuenzaniso
    • Zvinongokanganisa dhizaini yekupinza
  7. Why does sparse feature activity reduce interference when more features than dimensions are represented?
    • Sparse activity makes all feature directions orthogonal
    • It increases the number of neurons automatically
    • Fewer features are active at the same time, so overlapping directions collide less often
    • It removes the need for nonlinear computation
  8. What does an antipodal pair of feature directions look like geometrically?
    • Two vectors at a right angle in separate dimensions
    • Two vectors with identical directions and magnitudes
    • A vector with no coordinates
    • Two directions that point opposite ways along the same line