系統設計者的話：學英文根本不用背，你講中文，你有在背中文嗎？學英文也是一樣的道理，那是自然而然就會的東西，用背的方式最後肯定都是不行的。 更多...

最好的學習就是有英文的父母，不斷的跟你說話、耐心的教導你，時間長了，你也就慢慢就融入其中了。

但是我們沒有英文父母怎麼辦？看影片學英文就學的起來嗎？學不起來。

我們從小到大，也看過了不少英文電影，至少也看過1、200部也跑不掉，但是為什麼看了這麼多的英文電影，英文還是很爛呢？

主要的原因是我們沒有從基礎性的強化學習。舉個例子，西方的幼兒出生時，他至少聽了各種大人講英文至少聽了2年，才慢慢開始說英文、與大人互動。

而且都是一字一句慢慢的說、慢慢的教、漸進式的、慢慢的、越來越習慣、越來越聽的懂、越來越會說的過程。

那麼我們看的影片、電影過程中，都是大人成熟型的講話速度、方式、語調，我們怎麼可能跟的上？

一旦跟不上，思想中就會產生覺的好難哦、我不行種種念頭，那麼產生了這種潛在式的排斥心理作用，那麼你就更難學的會了。


所以我們必須打好基礎、深厚的基礎，萬丈高樓平地起，唯有打下深厚的基礎，後面你走的路才會越來越順。

程式界裏有句話，程式債欠的越多，最後就只能放棄。學英文也是一樣，英文單字債欠的越多，你就直接上來文法、句型、語調，最後也只能放棄。

基於上述的思維，我打造了這套系統，從最基礎的單字學習起，不斷的強化、加深你對這個單字的聽力，透過各式各樣的影片、各式各樣的人、各種不同的場景。

就只是為了訓練一個單字的聽力，無數的人跟你講同一個單字，講個20遍、50篇、100遍(且不重覆)，我就不信學不起來。


這套系統做出來後，我自己在使用的過程中，確實真的能聽出來他們講的這個單字，而且他們講的速度再快也能聽的出來。

能聽的出來，就知道他講的大概意思；能夠知道他講的大概意思，就能夠用幾個簡單的單字回應；能夠回應，也就能夠對話；當能夠對話，就能夠再修正文法、句型、語調。


我的目地就是希望自己至少能聽的懂人家在說什麼，用英文跟人家對話，大家共同努力、學習。若有其它想法或對這套系統設計的建議，也可以與我聯絡：ichich2013@gmail.com

Distributed TensorFlow (TensorFlow Dev Summit 2017)
觀看次數：811
so
as
zack
said
i'm
Derek
murray
i've
folks
on
the
livestream
who
can't
put
flow
well
it's
what
enables
us
here
at
tensorflow
on
lots
of
tutors
at
the
same
v3
model
scales
up
using
up
to
200
gpus
large
embedding
matrices
it
can
be
gigabytes
or
tens
of
gigabytes
in
size
ic
lr
2017
to
Train
neural
networks
with
techniques
like
conditional
computations
it's
great
if
you're
an
expert
user
but
the
low-level
time
to
flow
API
and
so
my
so
is
to
give
you
a
bottom-up
what
tangible
treats
the
remote
devices
the
runtime
will
put
the
variables
in
a
gradients
whereas
the
worker
replicas
pre-processing
the
forward
and
back
prop
multi-core
cpus
and
gpus
and
one
big
this
belief
and
in
tensorflow
explicitly
to
loosely
coupled
computations
or
we
could
have
the
training
the
forwards
in
the
backdrop
name
hopefully
become
self-explanatory
graft
replication
is
just
to
split
up
a
standard
like
tensor
flow
optimizer
and
a
big
modification
to
your
existing
replicas
and
the
client
gets
bogged
down
currently
what
we
use
most
internally
is
devices
are
local
to
that
worker
replica
independently
and
they
get
mapped
to
different
subsets
of
the
devices
that
so
each
replica
places
its
variables
on
variable
and
if
you're
doing
replicated
like
applying
its
gradients
will
be
visible
and
task
1
and
vice-versa
so
device
strings
and
you
know
they're
all
variables
or
distribute
the
networking
customize
the
placement
of
every
single
pre-canned
device
functions
to
make
your
these
variables
I
just
think
it's
the
functions
like
embedding
lookup
for
offload
some
of
the
embedding
lookup
computation
and
the
gradients
to
the
combinatorial
a
large
problem
domains
train
replica
device
setter
is
the
device
functions
that
override
the
you've
got
all
the
devices
set
I
know
it's
like
you're
really
just
a
kubernetes
or
maze
offs
will
do
this
for
spec
and
then
you're
good
to
go
task
is
much
simpler
of
threads
that's
what's
called
join
and
misconception
because
this
is
your
ps
attentively
works
programming
primitives
as
these
little
customize
the
optimization
algorithm
or
long-running
training
job
on
a
set
of
of
leslie
lamport
long-running
training
process
so
a
saver
equals
true
tells
tensorflow
right
the
because
that's
where
the
numbering
start
of
day
and
logging
summaries
for
checkpoint
model
and
evaluates
the
test
wispy
couple
coordination
a
bit
easier
certainly
effectively
stateless
so
when
crash's
as
a
ps
faster
though
that's
a
something
like
apache
zookeeper
or
xyg
fails
and
kind
of
they'll
failover
explicitly
initialize
the
variables
or
going
to
do
is
it
will
automatically
initialize
variables
or
by
restoring
the
initializers
if
you're
not
the
chief
customizations
here
so
we
support
these
installable
hooks
that
you
can
attach
sort
of
a
tastefully
curated
set
of
have
these
high-level
api's
that
have
there's
like
tf
contrib
levels
support
can't
dnn
classifier
with
two
hidden
boilerplate
high-level
API

相關影片

加Line好友 每日8AM 推播1單字

加入好友

其他人正在學習的單字

1. mongrel
2024-05-04 10:40
2. there
2024-05-04 10:40
3. spades
2024-05-04 10:40
4. site
2024-05-04 10:40
5. surprising
2024-05-04 10:39
6. green
2024-05-04 10:39
7. help
2024-05-04 10:39
8. clip
2024-05-04 10:39
9. shove
2024-05-04 10:39
10. stability
2024-05-04 10:39
11. love
2024-05-04 10:39
12. lovely
2024-05-04 10:39
13. thirtieth
2024-05-04 10:39
14. turns
2024-05-04 10:39
15. ago
2024-05-04 10:39

本月觀看影片單字排行榜

本月查閱單字排行榜

本月聆聽單字排行榜