![深度學(xué)習(xí)會(huì)引領(lǐng)到人工智能嗎_第1頁](http://file4.renrendoc.com/view14/M0B/1E/04/wKhkGWYm8RyAVOLTAABsaPTXjJQ797.jpg)
![深度學(xué)習(xí)會(huì)引領(lǐng)到人工智能嗎_第2頁](http://file4.renrendoc.com/view14/M0B/1E/04/wKhkGWYm8RyAVOLTAABsaPTXjJQ7972.jpg)
![深度學(xué)習(xí)會(huì)引領(lǐng)到人工智能嗎_第3頁](http://file4.renrendoc.com/view14/M0B/1E/04/wKhkGWYm8RyAVOLTAABsaPTXjJQ7973.jpg)
![深度學(xué)習(xí)會(huì)引領(lǐng)到人工智能嗎_第4頁](http://file4.renrendoc.com/view14/M0B/1E/04/wKhkGWYm8RyAVOLTAABsaPTXjJQ7974.jpg)
![深度學(xué)習(xí)會(huì)引領(lǐng)到人工智能嗎_第5頁](http://file4.renrendoc.com/view14/M0B/1E/04/wKhkGWYm8RyAVOLTAABsaPTXjJQ7975.jpg)
版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡介
Will
Deep
LearningLeadto
AI?Haytham
FayekML
/
AI
Melbourne
Meetup
-August
2017What
is
Artificial
Intelligence?What
is
Intelligence?Work
In
ProgressThe
ability
to
perceive
information,
andto
retain
it
as
knowledge
to
be
appliedtowards
adaptive
behaviors
within
anenvironment
or
context.3Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?What
is
Artificial
Intelligence?Think
Rationally Act
Rationally4Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?[*]
Images
from
variousonline
sourcesThink
Like
Human
Behave
Like
HumanAI5Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?LearningMeta-LearningPerceptionAttentionMemoryReasoningPlanningEmotionCommunicationNarrow
AIvsStrong
AIWhat
isDeep
Learning?The
hierarchical
learning
of
feature
representations6Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?[*]
Images
from
variousonline
sourcesNotationNotationLinear
Operation:Non-Linear
Operation:Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?8NotationAFully-Connected
Layer:A
Convolutional
Layer:Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?9NotationxhyHaytham
Fayek,
Will
Deep
Learning
Lead
to
AI?10NotationxhyFind
W
that
minimizes
L:Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?11Notation[*]
/neural-network-zoo/Feedforward
Neural
Network(DNN)Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?12Convolutional
Neural
Network(ConvNets)Recurrent
Neural
Network(RNN)Long
Short-term
Memory(LSTM-RNN)Residual
Network(ResNets)LearningLearningSupervised
LearningPick
a
model,
loss
function
&
training
scheme.Train
a
large
model
on
lots
of
labeled
data.[1]
Zeiler
&
Fergus,
Visualizing
and
Understanding
Convolutional
Neural
Networks,
2013Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?14LearningReinforcement
LearningHaytham
Fayek,
Will
Deep
Learning
Lead
to
AI?15LearningReinforcement
Learning[2]
Heesset
al.,
Emergenceof
LocomotionBehaviours
inRich
Environments,2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?16PerceptionPerceptionSupervised
Learning
has
made
strides
in
Perception:Speech
Recognition,
Computer
Vision,
Natural
Language
Processing.Each
modality
requires
a
fair
amount
of
specialized
engineering.18Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?[*]
Images
from
respectivepapers:
He
et
al.2015,
Hannun
et
al.
2014PerceptionA
single
model
with
good
results
onmultiple
tasks
in
multiple
domains.Tasks:
Object
Recognition,
translation,image
captioning,
speech
recognition,parsing,
..Components:Convolutional
LayersAttention
MechanismMixture
of
Experts[3]
Kaiser
et
al.,
One
Model
toLearnThem
All,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?19Perception[4]
Kaiser
et
al.,
One
Model
toLearnThem
All,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?20AttentionAttentionAllows
selective
processing
of
input
to
avoid
clutter
and
simplify
computation.Allows
processing
of
complex
input
in
nonlinear
ways.Attention
can
make
our
models
more
interpretable.[5]
Cho
et
al.,
Describing
Multimedia
Content
using
Attention
Based
Encoder
Decoder
Networks,
2015Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?22Attention[6]
Xu
et
al.,
Show,
Attend
and
Tell:
Neural
Image
Caption
Generationwith
Visual
Attention,
2015Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?23Image
Caption
GenerationMemoryMemoryNeural
Turing
MachineAnLSTM
ControllerRead
/
Write
Heads
usingAttention
(NN)An
External
MemoryCan
learning
simple
algorithmssuch
as
copying
&
sorting[7]
Graves
et
al.,
Neural
Turing
Machine,
201525Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?ReasoningReasoningProcess
very
unstructured
complex(multi-modal)
inputDeliberate
and
infer
relationsStore
KnowledgeOutput
unstructured
output[8]
Johnson
et
al.,
Learningand
Executing
Programs
for
Visual
Reasoning
,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?27ReasoningVisual
Question
Answering
(VQA)[9]
Yang
et
al.,StackedAttention
for
ImageQuestion
Answering,2015Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?28ReasoningVisual
Question
Answering
(VQA)[9]
Yang
et
al.,StackedAttention
for
ImageQuestion
Answering,2015Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?29ReasoningVisual
Question
Answering
(VQA)[8]
Johnson
et
al.,
Learningand
Executing
Programs
for
Visual
Reasoning
,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?30ReasoningVisual
Question
Answering
(VQA)[8]
Johnson
et
al.,
Learningand
Executing
Programs
for
Visual
Reasoning
,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?31Meta-LearningMeta-LearningLearning
to
Learn[*]
Ravi
et
al.,
Optimization
as
a
Model
for
Few-Shot
Learning,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?33Meta-LearningLearning
to
Learn[*]
Ravi
et
al.,
Optimization
as
a
Model
for
Few-Shot
Learning,
2017Haytham
Fayek,
Will
Deep
Learning
Lead
to
AI?34Meta-Lear
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。
最新文檔
- 2025年股東聯(lián)合健身房合作發(fā)展協(xié)議
- 2025年辦公室裝潢設(shè)計(jì)合同范本
- 2025年紅外線汽車尾氣分析儀項(xiàng)目提案報(bào)告模板
- 2025年企業(yè)股權(quán)交易合同批準(zhǔn)標(biāo)準(zhǔn)
- 2025年農(nóng)作物機(jī)械化種植產(chǎn)業(yè)鏈優(yōu)化合作協(xié)議
- 2025年噪聲污染治理項(xiàng)目申請報(bào)告模板
- 2025年協(xié)作創(chuàng)建美術(shù)培訓(xùn)中心合同樣本
- 2025年財(cái)產(chǎn)保全擔(dān)保合同范本
- 2025年農(nóng)產(chǎn)品冷藏儲(chǔ)存與物流配送服務(wù)協(xié)議書
- 2025年公路建設(shè)項(xiàng)目合同模板
- 報(bào)價(jià)單(報(bào)價(jià)單模板)
- 刑事案件模擬法庭劇本完整版五篇
- 2014教師事業(yè)單位工作人員年度考核登記表1
- 烏海周邊焦化企業(yè)概況
- Flash動(dòng)畫設(shè)計(jì)與制作(FlashCS6中文版)中職PPT完整全套教學(xué)課件
- Hadoop大數(shù)據(jù)開發(fā)實(shí)例教程高職PPT完整全套教學(xué)課件
- 新人教版小學(xué)數(shù)學(xué)五年級下冊教材分析課件
- 企業(yè)中層管理人員測評問題
- 人教版高中地理必修一全冊測試題(16份含答案)
- 水泥攪拌樁水灰比及漿液用量計(jì)算表(自動(dòng)計(jì)算)
- 教育:創(chuàng)造無限可能
評論
0/150
提交評論