Youtube daily report Dec 29 2017

Hello and welcome to Best Replays of the Week.

In this episode we have: a Blackdog making a mess,

an Invader Top Gun Steel Wall Conqueror, and an invisible Defender.

Roll intro!

First we go to the North American region with Bob40000

who's fighting a standard battle on Murovanka.

This Blackdog is the best Tier VIII light in the game after all the changes,

so let's see how it can operate in a tough Tier X matchup.

The eastern forest is a great place to use this light's camo and gun depression.

Even side on the Maus is a tough nut to crack.

But keep trying and eventually one does go in!

This ISU, however, is much more attractive!

But we must be careful, as the enemy's light tanks are still in hiding.

Looks like the team is in trouble as enemies try to cross the ridges,

but even if it's difficult to lay down effective fire,

the enemies are likely shaken.

This Conqueror is distracted, time to move in. Charge!

Watching these HESH rounds pen almost makes me feel bad for this Conq,

until I see all dat damage. Burn baby, burn!

That was an almost full HP Conq in just 4 pens.

But that's not enough!

Bob wants more, and this spotted Skorpion is just the thing.

Followed by a juicy big bootied British Tortoise!

Bob's teammate gets wrecked by the FV, revenge is needed! Om nom nom.

Eek! A brave reposition is in order as this Bat has ruined all the fun.

You're next on the hitlist, buddy, just you wait.

After some patient driving, Bob spots the Bat, and it's trying to shoot at allies!

Count the shells:

one,

two,

three, four, and that's all 5!

Go, go, go! You can run, but you cannot hide, Bat-Chat!

7,000 damage in close quarters in a Bulldog—that's no easy feat!

And the 2,000 base xp tells us just that.

Crucial? More like carry contribution! Well played, Bob40000!

Sticking with the North American region for now,

we meet Seras_I in their SU-122-44.

This TD has been neglected in recent years,

but that takes nothing away from its raw damage dealing potential.

Instead of camping for the whole battle, Seras heads to the hill,

often the deciding factor for an encounter on Fiery Salient.

First enemies are spotted in the middle

and this hill position provides adequate flank shots.

This TD has unnerfed old school camo values,

so spotting through this thin ridge of bushes yields a great result.

Even firing does not reveal Seras' position.

If only the penetration wasn't so ineffective against these heavy tanks.

However, I'd be more cautious with the ammo in this Soviet TD.

I'm often running out in battle.

Well if you can't pen, spot!

This IS' side is a much more palpable target than that T32.

It's like the floodgates have opened for damage here on the cap.

You couldn't dream up a better situation for this Russian DPM machine!

Boom party!

Seras managed to go from 800 damage to 3,700 in just over a minute.

But there's more to be done as allied numbers dwindle.

Expert use of the bushes here keeps the Leo at bay.

When was the last time this beast was spotted I wonder?

This should be the Ninja award!

This WZ is the biggest threat at the moment

but the shots aren't presenting themselves.

The two British TD's are dead in the water as the enemy team swarms forwards.

All that can be done from the hill

is to provide occasional shots through the buildings.

Some careless shooting leaves the Skorpion alive.

It also leaves the ammo empty, save for some HE rounds.

With 5 enemies remaining and the cap trundling forward, victory is impossible.

The Leo from the hill is the one to strike the final nail in the coffin.

Great usage of camo and spotting in this relic of a TD.

But this should be a lesson to all of us

that you need every skill to turn the tide of war.

Good try, Seras.

We venture now to the EU region with Levyelisapik on Lakeville.

The Conqueror in a 3-5-7 top tier match is a force to be reckoned with,

even more so in an engagement over a ridgeline.

With only a single arty the 2 line is actually a good position

to push as a top tier heavy like the Conqueror.

These meds don't fancy their chances peaking more

once they've seen this beast trundling forward.

As I thought, this damage farm is almost comically easy.

That Jagdpanther has had enough and backs off,

however the rest of their team hasn't got the message yet.

Let's hammer it home.

Emm, nice clip Škoda? Okay...

With one of the three Tier IXs out of the way,

it's time to mop up the last of this opposition.

Status check. Five minutes gone, 7 enemies left and, oh crap! Only 11 shells left.

I know the Conq starts off a bit short of shells

but this battle has been ridiculous!

These two TD's shouldn't pose much of a problem. Arty, on the other hand…

Sitting and getting clicked won't help anything,

so pushing and taking out these TD's is the best play to make.

RIP crew, and the med-kit is still on cooldown. Great…

Not much happens until the Commander is able to be patched up

and put back in action,

now it's time to draw in the remaining enemies with a cap.

The enemy WZ has made it back with ample time to reset.

I can't see this ending good.

A stroke of luck! For some unknown reason this WZ charges forward

and allows himself to be facehugged. Maybe there's a chance?

Or maybe not…

This WZ seems to be having similar problems.

This duel screams "I'm panicking!"

A boun-… oh wait that's a HESH!

There's the low ammo count of the Conq coming back to bite our hero.

It's all over, the allied arty took a justified risky shot,

but… the T/25! They did it!

Leaving our hero with 54 HP and a solitary HESH round.

And I've got a good idea who that will be for.

Sweet, sweet revenge.

Although, if Levyelisapik waited just a few seconds more it would've been a Fadins.

Great pressure on the enemy team throughout the whole battle.

They didn't get a moment's rest from this week's Invader, Levyelisapik.

What a battle!

Sticking with the EU region,

we drop down a few tiers to a Redshire 5-10 encounter.

Taking us on a tour will be XiteX who is on a quest for their 3rd mark.

This Type 64 is a monster when paired against its own tier

and Redshire also offers lots of bushes and ridges

to utilise the excellent camo of this little Chinese scout.

This aggressive bush offers perfect spotting opportunities

on the enemies main ridgeline.

Surprise, ELC! You certainly didn't expect to be taking shots there, did you?

A Churchill III's side is too juicy to pass up.

Trying to passive scout MT-25? Pay the price for sitting exposed for too long!

Burn baby, burn.

The fire eats away this Churchill's health like candy,

and a quick calculation tells our hero that there's no need to stay exposed.

The flames finish the job.

Two revenge shots from these enemies leave the little Type 64 on half HP.

Revenge is swift but it proves mistakes are instant in this machine.

Abusing the terrain quickly shuts this Panzer IV down.

Although this ridge is a good place to shoot from,

it's exposed when the flanks around it are lost

and that shot in the rear is a good reminder.

A Churchill and a T-34-85M? XiteX's surrounded. Keep calm.

An exchange of shots leaves XiteX with just 14 hit points!

Thankfully the T-34-85M is dealt with by an ally.

After a quick loop around the middle of the map

XiteX finds themselves right back under the bridge where they started.

Surrounded on 3 sides, perhaps the river can provide the means of escape.

Take every opportunity to do damage, then move!

Notice it's 1 versus 6 now?

Rapidly altering our location is the best way to keep the enemy team scattered.

Don't get distracted, use the clock, Luke! I mean XiteX.

Found you Hellcat! One shot one kill.

Some of the wiser reds have caught on and started a cap…

but not all, 3 others have continued their hunt.

Good shot!.. Not so good.

But this STRV fluffs their shot! Pop!

And this T67 loses its nerve and doesn't just yolo.

And loses the duel.

There's just over a minute left on the cap.

Lucky it's an encounter else that cap would've been a bigger problem!

There's one, what a reset!

Location, location, location as they say. And there's two!

Lawl, likely thought the battle was won and went AFK. Tut, tut.

1vs1 and two minutes on the clock.

There it is! And a beautiful pen puts the Churchill down to a one shot!

Oh, it went to cap did it? Interesting choice.

Relocate once more, and that's game!

Well done, XiteX, a 1v6 is never an easy thing to do,

but you did it without missing a step.

Achieving your 3rd mark and a Kolobanov's

hopefully makes our Defender award a little bit sweeter.

We have arrived at the pinnacle of the show!

EliteSoldier972 has volunteered to be our star from the North American region.

This Mäuschen is seldom spotted on the battlefield,

despite being a recent introduction.

Its gun and armor are not spectacular and it struggles at range.

However, EliteSoldier is going to prove to us

that the Mäuschen cannot be overlooked on a map such as Airfield.

Much of the fighting in the standard heavy area actually favours this type of tank.

Right away we get a good example of what not to do.

That's not how you peak, Caernarvon. Down boy!

OK, a bit of a strange start, but free damage is always nice, Victoria.

Whilst love tapping the Type 59,

a JagdTiger 88 plinks away at the Mäuschen's armor, to no avail.

This Type 59 really doesn't give up!

Hey, E-75, eyes on me!

A pen?! Wait, 21 damage from that base sniper?

This JagdTiger 88 is really throwing all it has at EliteSoldier this battle.

The Type 59 is rewarded for its persistence with a penetration at last.

And now arty has found its mark and joins the barrage from the JagdTiger 88.

Track gone and crew stunned. We're going nowhere fast.

Finally! The track is up long enough to destroy the pesky Type 59.

And why not this T-54, while we're at it!

That E-75's death marks the end of the flank. Good job everyone… oh.

Middle is lost and enemies have now swarmed up the centre road

so our hero must turn to engage. Hi, T32!

Forget about the JagdTiger 88, even from the rear it can't do anything.

Eh, nevermind that last comment,

it's time to retreat to the safety of the balcony. Where is that TD anyway?

There! Good job, IS-6!

Oh, and the arty! Very good job, IS-6!

Hello there Defender, if you insist.

This T28 Prototype was a bit too eager and got slapped for it's trouble.

A nicely placed hatch snipe

reduces the most dangerous tank on the enemy team to a smoking wreck.

The only ally that remains is an arty, and EliteSoldier is surrounded.

Surprise IS-6!

Oh and the JagdTiger! I'm going to enjoy this.

Good night, sweet JT. Good night.

Please form a neat line in front of the Mäuschen

and you will be killed in an orderly fashion. Thank you!

After such a massacre the IS-6 freezes in terror!

Ten kills. Now it's time to hunt this VK P and end this!

On the two minute warning and ugh, not the most optimal position to engage from

but let the final showdown commence!

Through the mantlet, now that's just showing off.

This VK hasn't realised that their tank is longer than it is wide

and is sidewards on, not advisable.

I tip my hat to you, EliteSoldier, for that monster of a carry.

More than two thousand base xp in Tier IX is a rare sight to see.

Nevermind the 11 kills and Kolobanov's to accompany it!

That certainly was the Best Replay of the Week.

That's all for this week.

Remember: some weeks we get better replays than others,

and if you think you got a better one, well, send it in!

I'm Luke Kneller, thanks for watching, and I'll see you next time.

For more infomation >> World of Tanks - Best Replays of the Week International #73 - Duration: 14:08.

-------------------------------------------

Jaguar S-Type 2.5 V6 EXECUTIVE Airco ECC Cruise control PDC 148dkm NAP Inruil mogelijk - Duration: 1:06.

For more infomation >> Jaguar S-Type 2.5 V6 EXECUTIVE Airco ECC Cruise control PDC 148dkm NAP Inruil mogelijk - Duration: 1:06.

-------------------------------------------

LE DOSSIER - Duration: 2:30.

For more infomation >> LE DOSSIER - Duration: 2:30.

-------------------------------------------

Here's the Time lapse Video of SpaceX's 'UFO' Rocket Launch - Duration: 3:49.

Here�s the Time-lapse Video of SpaceX�s �UFO� Rocket Launch That Sent the Internet

into a Frenzy

by Ivan,

An unusual beam of light illuminated the California sunset sky.

Many wondered if they were actually witnessing a real UFO sighting or even a nuclear attack

by North Korea.

But it was not about that.

The streets were filled with people looking at the sky amazed by what they were seeing.

Many drivers stopped their cars and left them to watch the show, whatever the cause was.

And in social networks, many Internet users showed their perplexity and curiosity.

One of them was Will.i.am, the singer of the popular group Black Eyed Peas, who published

a video with the mysterious lights and asked his thousands of followers: �What is that

in the sky?�

Many joked about the strange lights and recounted seeing them before on TV, as it�s not the

first time that something like it has been seen in the sky.

However, what millions of people that witnessed the strange sights did not know is that it

wasn�t a UFO. In fact, what they were seeing was a rocket from SpaceX, launched from the

Vandenberg Air Base, in Santa Barbara.

The Falcon 9 rocket was carrying ten satellites into orbit, bound for the Iridium constellation,

a belt of communications satellites orbiting Earth.

People started getting nervous about what they had seen.

It was the Fire Department of the city that, in the face of increasing alarm and stupefaction,

reassured the population with a communiqu�.

�It has been reported that the mysterious lights in the sky are the result of the launch

of a rocket from the Vandenberg base to put a satellite in space.�

The luminous strokes could be seen even in places of the state of Arizona.

Musk joked on Twitter while the residents of LA remained stunned.

Musk Tweeted: � Nuclear alien UFO from North Korea.�

Oh, here�s the video, enjoy:

Actually, a launch like Friday�s is nothing exceptional.

This year SpaceX has already carried out similar ones with different purposes.

The latest sighting was spectacular because of the time in the day it took place, as the

lightning conditions turned the rocket launch into a spectacular sighting.

However, while we are aware of the fact that Angelenos were fooled into a UFO sighing,

similar �sightings� have been recorded and photographed in the past.

This image for example was photographed during the launch of the Kosmos-1188 spacecraft on

June 14, 1980.

Similar to Musk�s rocket, right?

The following image, photographed by Leonard Lamoureux in 1937 shows what is officially

categorized as a UFO over Vancouver, British Columbia.

According to UFO Digest, ��the craft moved diagonally from the courthouse to the city

hall and hovered momentarily in front of Leonard. It sat in the air suspended above a flag pole��

Now if we are to delve into conspiracy waters and take a quick peek at the Ancient Astronaut

theory we find how the above images and videos of rocket launches resemble a UFO depicted

in a painting from the year 1350, dubbed �The Crucification,� which is located above the

altar at the Visoki Decani monastery in Kosovo.

For more infomation >> Here's the Time lapse Video of SpaceX's 'UFO' Rocket Launch - Duration: 3:49.

-------------------------------------------

This Is What The Moles On Your Face Mean - Duration: 4:13.

This Is What The Moles On Your Face Mean

First, check your face for any moles, and then look at the diagram above to identify

the number(s) that are the closest match to the moles on your face.

Usually, the moles only hold meaning for you if they are prominent and they are the only

one.

If your face if full of spots, acne or �little� moles, they do not count.

When you�ve ascertained which position corresponds to the mole on your face, look up the meanings

listed by numbers below.

POSITION 1 TO 3 As a child, you are somewhat rebellious and

a free spirit.

You have an innate creativity and work best when you are given a free hand.

Generally, your superiors like your avante garde approach to life.

If you have a mole here, you are far better off in business and being your own boss rather

than working for somebody.

What is promising is that you have the luck to be your own boss.

POSITION 4 You are an impulsive person, often acting

with a flamboyance that gives you charisma and a sparkling personality, but you can be

difficult when there are too many opinions.

You tend to be rather argumentative, but never to the point of holding grudges.

This mole tends to give you an explosive temper and should you decide to remove it, you will

find yourself becoming calmer and more at peace with the world.

POSITION 5 A mole above the eyebrow indicates that there

is wealth luck in your life, but you will need to earn it and work harder than most

people.

All the income you make must be carefully kept as there are people who are jealous of

you who might attempt to sweet talk you into parting with your wealth.

Be wary of those who try to interest you in get-rich-quick schemes.

If you have a mole here, it is advisable not to be too trusting of others.

Follow your instincts and be cautious.

And never allow other people to control your finances.

POSITION 6 A mole here indicates intelligence, creativity,

and skill as an artist.

Your artistic talent can bring you wealth, fame, and success.

It also indicates wealth luck, but this can only be fully realized if you follow your

heart rather than stick to conventional means of making a living.

Success will come if you are brave.

POSITION 7 Moles under the eyebrows indicate arguments

within the extended family that cause you grief and unhappiness.

This will affect your work and livelihood.

It is advisable to settle any differences you have with your relatives if you want peace

of mind to move ahead.

POSITION 8 This is not a very good position for a mole.

Your financial position will constantly be under strain because of a tendency to overspend.

You also have a penchant for gambling.

The only thing is you must know when to stop.

Meanwhile, someone with a mole here has a tendency to flirt with members of the opposite

sex as well as with the same sex.

Better be a little discerning where you exert your charms, or you might get into trouble.

POSITION 9 This mole position suggests sexual and other

problems.

It is an unfortunate mole and you are well advised to get rid of it.

It brings a litany of woes and a parade of problems.

POSITION 10 A mole here just under the nose indicates

excellent descendants luck.

You are surrounded by family at all times and will have many children and grandchildren.

You have the support of those close to you and will be both materially and emotionally

fulfilled.

POSITION 11 Moles here suggest a tendency to succumb to

illness.

It is a good idea to have this mole removed especially if it is a large, dark-coloured

mole.

Otherwise, use lots of foundation to cover it.

For more infomation >> This Is What The Moles On Your Face Mean - Duration: 4:13.

-------------------------------------------

MBC 연예대상 '대상 후보'에 자기 이름 불리자 깜짝 놀란 박나래 - Duration: 2:47.

For more infomation >> MBC 연예대상 '대상 후보'에 자기 이름 불리자 깜짝 놀란 박나래 - Duration: 2:47.

-------------------------------------------

[REFLEXOLOGIE PLANTAIRE POUR ENFANTS] COMMENT DETENDRE ET CALMER UN ENFANT POUR MIEUX DORMIR - Duration: 3:52.

For more infomation >> [REFLEXOLOGIE PLANTAIRE POUR ENFANTS] COMMENT DETENDRE ET CALMER UN ENFANT POUR MIEUX DORMIR - Duration: 3:52.

-------------------------------------------

Fox News, Sean Hannity Both Grab Big Wins to Mark the End of the Year - Duration: 2:15.

For more infomation >> Fox News, Sean Hannity Both Grab Big Wins to Mark the End of the Year - Duration: 2:15.

-------------------------------------------

Weihnachtsmann auf Verbrecherjagd nach Fahrerflucht | Motorrad Nachrichten - Duration: 2:54.

For more infomation >> Weihnachtsmann auf Verbrecherjagd nach Fahrerflucht | Motorrad Nachrichten - Duration: 2:54.

-------------------------------------------

Migas de tortilla con huevo (desayuno) - Duration: 4:50.

For more infomation >> Migas de tortilla con huevo (desayuno) - Duration: 4:50.

-------------------------------------------

L'HISTOIRE DE TIMAL | UN RAPPEUR NÉE POUR FAIRE DU SALE - Duration: 11:50.

Hi everyone, it's NGZ and we find it for a new video

So in the previous video we talked about Badjer, in this video we will talk about Timal

Before I start this video I would like to thank everyone like, comment and share my videos

in the previous video I asked you to give me names of rappers that I could present later, you have been many to do and I really want to thank you

For those who do not know, Timal is a West Indian rapper He was born in 93 Saint-Denis more precisely

and grew up in Champs-sur-Marne as rapper Badger

Also for those who do not know Badger, do not hesitate to go see my previous video

In short, let's go back to Timal, he is now 20 years old, and many people have known him thanks to his freestyle series 'report' on the official chain of Daymolition

The first report was released on March 25, 2016

is currently has just over 500,000 views

3 months later he went on with the second report, which this time exploded, that is to say that this freestyle did not do 1,2,3 but 6 million views

and besides, I think he is part of the freestyles that we enormously help to make known, because personally, I knew him thanks to that the

Following the success of his previous freestyle, 1 month later Timal released the third report, which this time exceeded the first million, soon the second, it's just a matter of time

For more infomation >> L'HISTOIRE DE TIMAL | UN RAPPEUR NÉE POUR FAIRE DU SALE - Duration: 11:50.

-------------------------------------------

4 HEURES Bruit Blanc Ventilateur De Plafond Très Apaisante Pour Dormir- Bruit Blanc Relaxant - Duration: 4:03:44.

For more infomation >> 4 HEURES Bruit Blanc Ventilateur De Plafond Très Apaisante Pour Dormir- Bruit Blanc Relaxant - Duration: 4:03:44.

-------------------------------------------

Apple se kaje. S omluvou rapidně snižuje ceny pozáruční výměny baterie - Duration: 3:06.

For more infomation >> Apple se kaje. S omluvou rapidně snižuje ceny pozáruční výměny baterie - Duration: 3:06.

-------------------------------------------

How I Make Money Online

For more infomation >> How I Make Money Online

-------------------------------------------

Demain nous appar­tient : qui est Juliette, la femme d'Alexandre Bras­seur, sa compagne depuis - Duration: 2:38.

For more infomation >> Demain nous appar­tient : qui est Juliette, la femme d'Alexandre Bras­seur, sa compagne depuis - Duration: 2:38.

-------------------------------------------

BMW X3 2.0D XDRIVE HIGH EXECUTIVE AUT8/SPORTSTOELEN/NAVO PROFF/LEDER/TREKHAAK/19" - Duration: 1:01.

For more infomation >> BMW X3 2.0D XDRIVE HIGH EXECUTIVE AUT8/SPORTSTOELEN/NAVO PROFF/LEDER/TREKHAAK/19" - Duration: 1:01.

-------------------------------------------

Mercedes-Benz GL-Klasse GL 350 d 4MATIC Automaat | Luchtvering | Alarm | Xenon | COMAND - Duration: 0:52.

For more infomation >> Mercedes-Benz GL-Klasse GL 350 d 4MATIC Automaat | Luchtvering | Alarm | Xenon | COMAND - Duration: 0:52.

-------------------------------------------

Mitsubishi Colt 1.5 DI-D Cruise MOTOR DEFECT BJ 2006 !!! - Duration: 0:59.

For more infomation >> Mitsubishi Colt 1.5 DI-D Cruise MOTOR DEFECT BJ 2006 !!! - Duration: 0:59.

-------------------------------------------

Mazda 6 Sportbreak 2.2D Skylease+ Navigatie Airco, Cruise Control, Navigatie | RIJKLAARPRIJS | - Duration: 1:01.

For more infomation >> Mazda 6 Sportbreak 2.2D Skylease+ Navigatie Airco, Cruise Control, Navigatie | RIJKLAARPRIJS | - Duration: 1:01.

-------------------------------------------

Synthetic Gradients Tutorial - How to Speed Up Deep Learning Training - Duration: 20:25.

Hi, I'm Aurélien Géron, and today I'm going to explain how Synthetic Gradients can

dramatically speed up training of deep neural networks, and even often improve their performance

significantly.

We will also see how they can help recurrent neural networks learn long term patterns in

your data, and more.

Synthetic Gradients were introduced in a paper called "Decoupled Neural Interfaces using

Synthetic Gradients" published on Arxiv in 2016 by Max Jaderberg and other DeepMind

researchers.

As always, I'll put all the links in the video description below.

To explain Synthetic Gradients, let's start with a super quick refresher on Backpropagation.

Here's a simple feedforward neural network that we want to train using backpropagation.

Each training iteration has two phases.

First, the Forward phase: we send the inputs X to the first hidden layer, which computes

its outputs h1 using its parameters theta1, and so on up to the output layer, and finally

we compute the loss by comparing the network's outputs and the labels.

Then the Backward phase.

The algorithm first computes delta3, which are the gradients of the loss with regards

to h3, then these gradients are propagated backwards through the network, until we reach

the first hidden layer.

The final step of Backpropagation uses the gradients we have computed to tweak the parameters

in the direction that will reduce the loss.

This is the gradient descent step.

Okay, that's it for Backpropagation.

Now suppose you want to speed up training.

You buy 3 GPU cards, and you split the neural network in three parts, with each part running

on a different GPU.

This is called model parallelism.

Unfortunately, because of how Backpropagation works, model parallelism is inefficient.

Indeed, to compute the loss, you first need to do a full forward pass sequentially.

Each GPU has to wait for the previous GPU to finish working on a training batch before

it can start working on it.

This is called the Forward Lock.

Notice that the model parameters cannot be updated before the loss is computed.

And this is called the Update Lock.

And finally, we cannot update a layer's parameters before the backward pass is complete,

at least down to the layer we want to update.

This is called the Backward lock.

The consequence of all these locks is that GPUs will spend most of their time waiting

for the other GPUs.

As a result, training on 3 GPUs using model parallelism is actually slower than training

on a single GPU.

So, the main idea behind Synthetic Gradients is to break these locks, in order to make

model parallelism actually work.

Let's see how.

First we send the inputs to the first hidden layer.

Then this layer uses its parameters theta1 to compute its outputs.

So far, nothing has changed.

But now we also send the outputs h1 to a magical little module M1, called a Synthetic Gradient

model.

We'll see how it works in a few minutes, but for now it's just a black box.

This model tries to predict what the gradients for the first hidden layer will be.

It outputs the synthetic gradients delta1 hat, which are an approximation of the true

gradients delta1.

Using these synthetic gradients, we can immediately perform a gradient descent step to update

the parameters theta1, no need to wait.

This hidden layer equipped with its Synthetic Gradient model is effectively decoupled from

the rest of the network.

This is called a Decoupled Neural Interface, or DNI.

In parallel, the second layer can do the same thing.

It uses a second Synthetic Gradient model M2 to predict what the gradients will be for

the second hidden layer.

And it performs a gradient descent step.

And so on up to the output layer.

This time instead of using a Synthetic Gradient model, we might as well compute the true gradients

directly and use these true gradients delta3 to update the parameters theta3.

And we are done!

Notice that we only did a forward pass, no backward pass.

So just like that, training could potentially be up to twice faster.

Just to be clear, the Synthetic Gradient models are only used during training.

After training, we can use the neural network as usual, based on the trained parameters

theta1, 2 and 3.Okay, now let's see how this technique enables model parallelism during

training.

Once again, let's split the network into three parts, each running on a different GPU

card.

And the CPU will take care of loading the training instances and pushing them into a

training queue.

We start by loading the first training batch.

And while the first GPU is computing h1, and updating its parameters using synthetic gradients,

we can already load batch number 2 and push it into the queue.

Then while layer 2 takes care of batch number 1, layer 1 can already take care of batch

number 2.

No need to wait!

And so on, so you get the picture.

Now each layer is working in parallel on a different batch, so all GPUs are active, they

are much less blocked waiting for other GPUs to finish their jobs.

And we can continue like this until the end of training.

As you can imagine, this can dramatically reduce training time.

However, every time we go from one layer to the next, we need to move a lot of data across

the GPU cards.

This can take a lot of time and in practice it can far outweigh the benefits of this architecture.

But if you have a deep neural network composed of, say, 30 layers then you can split it in

3 parts of 10 layers each.

You can use Synthetic Gradient models at every hidden layer, or every few hidden layers,

or just at the interfaces between the GPU cards.

With so many layers, the time required to copy the data across GPU cards is now small

compared to the total computation time, so the GPU cards spend much less time waiting

for data, and you can hope to train your network close to 3 times faster than using regular

Backpropagation on a single GPU card.

So model parallelism actually works!

Great!

Now it's time to open the black boxes and see how the Synthetic Gradient models work.

Let's focus on a hidden layer i.

It has its own Synthetic Gradient model Mi which produces synthetic gradients delta i

hat, and these synthetic gradients can be used to update the hidden layer's parameters

without waiting for the true gradients to be computed, as we have just seen.

This model can simply be a small neural network.

For example, a single linear layer, with no activation function.

Or it could have a hidden layer or two.

We will simply train the Synthetic Gradient model Mi so that it gradually learns to correctly

predict the true gradients delta i.

For this, we can just train the Synthetic Gradient model normally, by minimizing a loss

function.

We can just use regular Backpropagation here, nothing fancy.

For example, we can minimize the distance between the synthetic gradients and the true

gradients (in other words, the L2 norm of their difference), or we can minimize the

square of that distance.

But this begs the question: how do we compute the true gradients delta i?

If we need to wait for the loss function to be computed and for the true gradients to

flow backward through the network, then we have somewhat defeated the purpose of synthetic

gradients.

Fortunately, there's a neat trick to avoid this.

We can just wait for the next layer to compute its synthetic gradients delta i+1 hat and

then we just Backpropagate these synthetic gradients through layer i+1.

This does not really give us the true gradients delta i, but hopefully something pretty close.

Of course if the next layer happens to be the output layer, then we might as well compute

the true gradients and Backpropagate them.

Over time, the Synthetic Gradient models will get better and better at predicting the true

gradients, and this will be useful both for updating the parameters correctly and also

for providing accurate gradients to train the Synthetic Gradient models in lower layers.And

that's it, you now know what synthetic gradients are, how they work and how they can speed

up neural network training.

But there are a few more important things to mention.

Firstly, Synthetic Gradients can be used pretty much on any type of network, including convolutional

neural networks such as this one.

Just add Synthetic Gradient models after some hidden layers, and that's about it.

Each Synthetic Gradient model's outputs must have the same shape as its inputs, that

is the same shape as the outputs of the layer they are attached to.

For example, M1's outputs must have the same shape as the outputs of this convolutional

layer.

Suppose it's a convolutional layer with 5 feature maps of size 400x200, then that's

exactly the shape that M1 must output.

That's a 5x400x200 array.

In practice, you can use a shallow convolutional neural network that preserves the shape of

its inputs, so for example a couple convolutional layers with zero padding and stride 1 would

do just fine.

Here's another important point.

Until now, the input of each Synthetic Gradient model Mi was only the output of the corresponding

layer, hi.

But it is perfectly legal to provide additional information to the Synthetic Gradient model,

so that it can make better predictions.

For example, we can give it the labels of the current batch.

This is called a conditional Decoupled Neural Interface, or cDNI.

In the paper, the authors show that cDNI consistently performs better than regular DNI, so it should

probably be your default choice.

So in the paper, they experimented with the MNIST dataset of handwritten digits, using

various architectures and training methods.

In particular, they used this fully connected network with 3 to 6 hidden layers of 256 neurons

each.

They used Batch normalization and the ReLU activation function at each hidden layer.

And here is a graph presented in Figure 2 in the paper.

It shows the learning curves for 3 to 6 hidden layers and for various training methods.

For example, when trained using regular Backpropagation, the network reaches below 2% error on the

test set, and it gets better when you add more layers.

Using Synthetic Gradient models at each hidden layer, the final performance of the 3 layer

network ends up being better than before, but it takes time to train the synthetic models,

so overall, you know, it's a little bit longer than Backpropagation.

When you add more layers, the network's performance actually decreases, and training

time increases.

That's not great.

Note that each synthetic gradient model is actually composed of two hidden layers of

1024 neurons each, and one output layer of 256 neurons.

They also used batch normalization and the ReLU activation function in the hidden layers.

Finally, they tried training the network using conditional DNI.

The network gets better when you add more layers, and with 6 layers it actually reaches

the best performance overall.

Moreover, as you can see, this is the fastest learning architecture.

It reaches less than 2% error in just a few thousand iterations.

Surprisingly, they used very simple synthetic gradient models, without any hidden layers

here.

I am curious to know why they did not use the same synthetic models for DNI and cDNI,

because it feels like we are comparing apples and oranges.

Anyway, it clearly demonstrates that cDNI performs much better than Backpropagation

on this task, both in terms of final accuracy and training speed.

There are many more results in the paper, if you're interested, in particular great

results with Convolutional Neural Nets.

Another great application of Synthetic Gradients is in Recurrent Neural Networks.

At each time step t, a recurrent layer takes the inputs Xt, as well as its own outputs

from the previous time step h_t-1, and it produces the output h_t.

It is convenient to represent RNNs by unrolling them through time, across the horizontal axis,

like this.

First the recurrent layer takes the inputs at time t=0, and it has no previous outputs.

It then outputs h_t=0

And at the next time step, it takes the inputs X_t=1 and the previous outputs h_t=0.

To be clear, these two boxes represent the same recurrent layer at two points in time.

Then it outputs h_t=1

And we could go on and on and on…

However, during training, we have to stop at one point, or else we will run out of memory.

We can then compute the loss based on the outputs produced so far.

And we can perform Backpropagation.

And finally we can update the parameters of the recurrent layer.

This technique is called Truncated Backpropagation through time.

It works well, but it has its limits.

In particular, since we only computed the loss on a few outputs, we know nothing about

the future losses.

So in practice, this means that the network cannot learn long-term patterns.

So let's see how Synthetic Gradients can help solve this problem.

Instead of stopping at time step t=3, let's unroll the network for just one additional

time step.

But instead of using its outputs to compute the loss, we send them to a Synthetic Gradient

model.

It estimates the gradients for that time step, delta_t=4_hat.

And we backpropagate these gradients through the layer to get an estimate of delta_t=3.

We can then perform regular Backpropagation through time, by mixing the true gradients

and the estimated future gradients.

Finally, once we have all the gradients we need, we can update the parameters of the

recurrent layer by performing a gradient descent step.

We must not touch the last unrolled cell, because this would change its output h_t=4,

and we are going to need it in a minute to train the Synthetic Gradient model.So by using

Synthetic Gradients in a recurrent neural network like this, we can capture long term

patterns in the data even if we unroll the network through just a few time steps.

Now, let's see how we can train the Synthetic Gradient model.

For this, we will need to run the network on the next few time steps, so let's move

forward in time.

Okay, clean up a bit and push this to the left to have more space.

Okay, now we run the RNN on the next few time steps.

Okay, we compute the loss.

We add an extra time step and we use the Synthetic Gradient model to estimate the gradients for

that time step.

And just like earlier, we Backpropagate these synthetic gradients and we mix them with true

gradients.

And now this process gives us something pretty close to the true gradients for time step

4, and we can use these gradients to train the Synthetic Gradient model.

Next, we can use the gradients we computed to update the RNN's parameters.

And boom!

Of course we could repeat this process many times, and both the RNN and the Synthetic

Gradient model would get better and better.It does add some complexity, but you can bet

that the main Deep Learning libraries will soon hide this complexity from us, hopefully.

And if you need some motivation, here are some amazing results.

This graph is a simplified version of Figure 4 in the paper, and it comes from DeepMind's

great blog post about Synthetic Gradients, which I highly encourage you to read (the

link is in the video description below).

It shows the performance of various RNNs on the Penn Treebank task, which is a language

modelling task.

The horizontal axis shows training time, and the vertical axis shows the model's error,

measured in bits per character (BPC).

The three dashed lines are the learning curves of a regular RNN using Backpropagation through

time, unrolled through 8, 20 or 40 time steps.

So the more you unroll the RNN, the longer it takes to train, and the more data it requires,

but also the better the performance it eventually reaches.

Now compare these three dashed lines to the solid line on the left: it shows the learning

curve of an RNN trained using Backpropagation through time unrolled through just 8 time

steps, but this time using synthetic gradients.

As you can see, the model reaches the lowest error, even better than the model unrolled

through 40 time steps, and it takes roughly half as much time and data to train.

That's really impressive!

Okay next!

Yet another really interesting idea in the paper aims to break the forward lock.

Recall that the Forward lock is the fact that we need to wait for the lower layers to finish

before we can compute the top layers.

It may sound impossible to break this lock, but it is in fact quite simple: you can just

equip any layer you want with a Synthetic Input model.

For example, let's add a Synthetic Input model I3 to layer 3, which is the output layer.

It allows us to skip the hidden layers 1 and 2 by computing h2_hat, an approximation of

h2, the inputs of layer 3.

We can just feed h2_hat directly to the output layer.

And ta-da!

We've just broken the forward lock.

As you might guess, once we eventually get the output of the hidden layer 2 we can use

it to train the Synthetic Input model.

This is really the exact same idea as earlier, but going forwards rather than backwards.

In fact, we can even use the same trick as earlier to go even faster.

Instead of letting the signal propagate through the whole network to compute h2, we can just

use the synthetic input model from the previous layer and feed it to the hidden layer 2, and

this will give us something hopefully close enough to h2, to train I3, the Synthetic Input

model of layer 3.

To conclude, let's look at the data flow of a fully Decoupled Neural Interface that

uses both synthetic inputs and synthetic gradients.

First, the Synthetic Input model receives the next training batch and computes an approximation

of the layer's inputs, h_i-1_hat.

Then, the hidden layer computes its outputs h_i and feeds them simultaneously to the next

layer and to its own Synthetic Gradient model.

These gradients are backpropagated through the hidden layer, which gives a reasonably

good approximation of the true gradients for the previous layer.

The gradients delta_i-1 are just sent back to the previous layer, which will use them

to update its own Synthetic Gradients model.

And immediately after that, we can update the layer's parameters using the Synthetic

Gradients delta_i_hat.

At some point we receive the outputs of the previous layer, h_i-1, and we will use them

to train the Synthetic Input model.

And lastly, we receive the gradients from the next layer, and we use them to train the

Synthetic Gradients model.

And that's it!

The DNI is ready to handle the next training batch.

If you want to learn more about Synthetic Gradients, I encourage you to read the paper

itself, as it touches on a few more topics, such as many implementation details, or how

Synthetic Gradients can help two Recurrent Nets communicate efficiently when they don't

tick at the same rate, and so on.

Also check out the links in the video description, there are several interesting blog posts and

implementations, and I might add my own implementation at one point.

If you want to learn more about Deep Learning, check out my book Hands-On Machine Learning

with Scikit-Learn and TensorFlow.

In particular, there's a whole chapter on running TensorFlow across multiple GPUs and

servers.

There's also a german version and a French version, and I believe a Chinese version should

be out in the next few weeks.

And that's all I had for today!!

I hope you enjoyed this video and that you found it useful.

If you did, please, like, share, comment, subscribe, and you can also follow me on Twitter

if you're into that.

See you next time and I wish you a very Happy New Year!

For more infomation >> Synthetic Gradients Tutorial - How to Speed Up Deep Learning Training - Duration: 20:25.

-------------------------------------------

Le coup monté d'Emma­nuel Macron et Cyril Hanouna pour Touche Pas à Mon Poste |Nouvelles générales - Duration: 3:06.

For more infomation >> Le coup monté d'Emma­nuel Macron et Cyril Hanouna pour Touche Pas à Mon Poste |Nouvelles générales - Duration: 3:06.

-------------------------------------------

7 INFORMATIONS INÉDITES QUE LA NASA NOUS CACHE - Duration: 10:07.

For more infomation >> 7 INFORMATIONS INÉDITES QUE LA NASA NOUS CACHE - Duration: 10:07.

-------------------------------------------

How Can I Get Faster? | Ask GCN Anything About Cycling - Duration: 8:43.

(soft pop music)

- As we inch ever closer

to the chequered flag of 2017, we're here, yet again,

for another edition of Ask GCN Anything,

so let's get the party started.

It's actually one of my all-time favourite questions

here on Ask GCN, and it comes from Karthik Sharama,

who asks, how to cycle very fast?

Well first off, thanks for getting in contact,

but here is my very simple answer.

One, get your bike.

Two, make sure it's in tip-top shape,

so clean, mechanically sound,

with the tyres pumped up,

and three, get out and ride your bike.

Four, enjoy.

Five, do that as often as you can and you'll get

fitter and fitter, and faster and faster.

And six, watch this video, which you could think

we made just for you.

- If you're not a particularly strong climber,

you don't have to go hard up the hills.

Instead, why not play to your strengths and go hard

over the top of climbs and on descents as well.

It's easy to focus on fitness as the key

to increasing your speed,

but don't fall into the trap of neglecting your skills.

It's almost irrelevant how fast you can go up a hill

if you can't go around corners and get yourself down

the other side quickly.

Practising these skills, like cornering and descending,

can really pay dividends.

Especially if you're lacking in confidence.

- It's time for the rapid fire round.

I'm gonna rattle these of as quickly as I can.

First up is this from Casey.

Do I have to worry about bike light mounts,

mainly the rear one, scratching my bike.

I'm worried the grit will get under

the rubber light mount clamp and the vibrations

from the road will scratch the seatpost or frame.

My advice, Casey, every time you've been for a ride,

clean it, replace it, job done.

Next up is this from Trevor Holmes.

Hi chaps, great content as ever

and heart warming vids from Dan.

Thanks very much.

My question is, is there a release date

for the cook book from the chef

that Si has been with recently.

Yes there is, it's out now.

You can get Hayden Groves' three tour cookbook

by heading to 3tourchallenges.com and you can buy it there.

And it's three, with a number three, not a written three.

Next up is this from Charles Farrell,

Hey guys, how come the riders

in the pro peloton use paper race numbers?

The teams spend so much on R and D for aerodynamic gains,

yet put paper numbers on the back of the jerseys.

Would it be more aero to use iron-on transfers like those

that are used to put team names on leaders jerseys?

Charles, I could not agree more.

You are right on the money.

Next up is this from Gareth Willox.

If you're heading out for a winter ride, not race,

about 100 K, is it a good or bad idea to warm up

on the trainer first, cheers.

Well Gareth, it's not a bad idea to warm up on the trainer,

there's no harm in doing it at all.

Or you could just warm up gradually out on the road,

but basically, do it if you feel

it's gonna be a little bit better for you.

Next up is this from L L.

We need another presenter challenge if you agree.

There's one coming up very very soon so watch this space.

Rod GU asks what's the usual weight gain

in kilogrammes during the offseason?

E G from the pros to you guys.

Some guys put on as much as five or six kilogrammes,

other guys don't put on much at all.

Generally speaking in the winter I used to put on

about two, two and a half kilogrammes but not much more.

But again, it depends on the physiology of the pro and also

how determined they are to keep the weight the same.

If you put too much weight on,

it can make it very difficult in the early part

of the season to get back to race weight.

But me, about two and a half kilos, no more than that.

And finally this from Olivier Picard.

How long should I wait before training again after a cold

to avoid getting sick again.

Best thing you can do is to listen to your heart

and also be honest about how you feel.

So if your heart rate is above 10 or 15 beats above

its usual rate then that's a pretty good warning signal

from the body to back things off.

So listen to your body, make sure you feel right,

and just for safety's sake, add another day on

as a bit of a buffer and start training nice and gradual.

Well let's slow things down just a little bit

with this question from fo-ad at-ah,

I hope I pronounced your name right there.

Hello GCN, now we all know that the flatter the terrain,

the more important the absolute power becomes,

and the hillier the terrain, the more important

power-to-weight ratio becomes.

Now since I live in a flat city

with maximum category four climbs,

should I hit the gym and get big strong legs

for me to generate more power?

Or should I be happy with my 61 kilos

and 4.5 watts per kilo cat four climbing skills.

Well Fouad, if you're pretty confident you're not

gonna be climbing anything bigger than a cat four,

then you are exactly right, absolute power

is the be all end all in essentially flat terrain.

Just make sure you don't become muscle bound

and put on so much weight that you actually

become pretty inefficient on even the smallest of climbs.

But yeah, get to the gym, work on your power,

work on your absolute power very much like lots

of the professional riders do these days.

Many many pros, especially sprinters who need to generate

those big amounts of power, they do get to the gym

and push some pretty heavy weights.

Or, if you're not really a big fan of the gym,

you wanna do something slightly different,

we've got some body weight workouts here that Simon

is very ably demonstrating in this video.

- Getting better at cycling isn't just always

about riding your bike.

Paul here, our resident fitness expert, is gonna take us

through a six minute workout that's gonna make us

fitter, stronger, and faster.

And Lord knows, Paul, I certainly need plenty

of that middle one.

- [Paul] Exercise number one is gonna be mountain climbers.

Exercise number two is gonna be squat lunge lunge.

Exercise number three is a core exercise

which is a ninety degree ab crunch.

Exercise number four is going to be five jumping jacks,

and then five floor jacks.

Exercise five is squat jumps,

and exercise number six is a straight arm plank,

with a little twist.

Finally on this weeks show, we have this question

from a jolly giant who asks,

I'm wanting to improve my FTP this winter.

FTP, remember, is functional threshold power,

how much power you can produce for an hour.

Now I've been cycling for six years doing 6,000 miles

a year or seven to eight hours per week.

And currently I have an FTP of just under 400 watts.

What percentage improvement should I

be looking at to achieve?

Thank you in advance.

Well first up jolly giant, thanks for getting in contact,

but also that is a mightily impressive

functional threshold power if your numbers are correct.

Now, if you're a heavier person, then an FTP

of around 400 isn't quite as unsurprising,

but by any stretch that is quality, so well done.

Now, if you're looking at making some improvements,

it's very difficult for me to put a percentage on it.

But just looking at how many hours you spend

on a bike per week and how many miles you do a year,

it's not out of the ordinary to expect you could do

a little bit more in terms of volume,

but more importantly, look at improving how smart you train.

So use that time wisely and I'm sure that you can find

some percentage gains there.

But what I think is vitally important here,

don't just focus on your FTP,

focus on other areas of your cycling.

So your endurance, your explosive power,

and also your top end speed as well.

And also your power-to-weight ratio.

If you keep your FTP the same, but drop a little weight,

basically you'll become far more efficient as a bike rider.

So, as I said before, very difficult to put

an actual percentage on this, it's just gonna depend

on how determined you are and how hard that you train.

But if it is your FTP you're particularly looking

at to improve, watch this video, I think it will help.

- Doing two 20 minute intervals or two 30 minute intervals

a couple times a week will do you a world of good.

- We've said it before but we'll say it again,

consistency is absolutely key

to making improvements in cycling.

So that means regular shorter workouts are more effective

than one massive ride per week.

- That's it for another edition of Ask GCN.

Please do keep those questions coming using

the #torqueback down in the comments section

and across on social media as well.

And we'll do our very best to answer

as many questions as we can.

Now, if you haven't already subscribed

to the Global Cycling Network you can do so for free

by clicking on the globe.

That way, you will not miss another video.

Now for a couple of other videos I think you will enjoy,

how about clicking just down here.

When Dan went to Zambia to basically look

at how much difference a buffalo bite makes.

Now that is an absolutely amazing video,

please cast your eyes and donate, watch that one.

And if you want something a little bit different,

how about clicking just down here for four more bits

of weird and wonderful retro tech.

For more infomation >> How Can I Get Faster? | Ask GCN Anything About Cycling - Duration: 8:43.

-------------------------------------------

LE DOSSIER - Duration: 2:30.

For more infomation >> LE DOSSIER - Duration: 2:30.

-------------------------------------------

Uomini e donne Trono Over: Giorgio contro la Galgani? | WInd Zuiden - Duration: 3:50.

For more infomation >> Uomini e donne Trono Over: Giorgio contro la Galgani? | WInd Zuiden - Duration: 3:50.

-------------------------------------------

5 Core Exercises To Make Yourself Stronger | Strength And Conditioning For Triathletes - Duration: 4:35.

- Core exercises tend to come in and out of fashion,

but I think by now we all realise

that actually it's a very important part of training.

Doesn't just make you more efficient,

but it helps the injury prevention,

especially if you're training for triathlons.

Though it's a good idea to strengthen this area

before a problem occurs.

But if you're anything like me,

it's one of those exercise that often comes

to the bottom of the list when you're trying to balance

three sports and everyday life.

So to get you started I've chosen five of my favourite

core exercises.

(upbeat music)

Side plank rotations are the first part of this.

It's important to get a really good solid side plank.

So I'm gonna use my forearm and elbow and lie that

on the mat at 90 degrees to my body.

And then gonna lift my hips up.

So the endpoint of contact will be my forearm and my feet.

Can either have one foot in front of the other like so.

I personally prefer it with one foot on top

of the other like this.

Now you should be completely at 90 degrees to the ground,

so facing a wall if you were looking at one.

From here this is where the rotation part starts,

so with your upper arm, in my case my left,

I'm gonna touch through behind my elbow,

touch the ground,

reach up to the sky.

Follow your hand as well so you rotate your head

and it just puts a little bit more challenge

on the rotation.

Back down, all the way up.

First time just try and do five of these.

Concentrate on doing them really well rather than

lots of them,

and then obviously you're gonna rest

and swap to the other side.

And I want you to try and manage three sets

for the first time.

(upbeat music)

A common misconception with core is it's about

having a six-pack or it's just working the front

of your stomach.

But actually it's the whole band around

this middle part of your body,

including your glute muscles.

And this exercise is gonna work your glutes.

And if you're anything like me,

they might be a bit lazy and they need a firing up.

So lying with your knees bent, on your back,

90 degrees roughly here.

Looking straight up,

and the important part is squeeze your glutes

and lift your hips up towards the sky.

Making sure you have a straight line from your knees,

hips, to your shoulders.

And then take your arms off and you've got the control,

and you want to do this nice and slowly

so you're focusing on getting it the correct movement,

10 times, then have a rest.

And do that three sets.

(upbeat music)

This exercise is as it sounds.

You're on your front.

You're gonna be walking out with your hands.

So to start you need to get into a press-up position,

little bit like a plank, but on your hands,

keeping everything nice and straight and controlled.

From here, you're gonna walk out with your hands,

taking small steps until you get to the point

where you feel you can just still hold it,

and then from there small steps back in

until your hands are underneath your shoulders.

Make sure you alternate with which side you start

just to keep things even.

And then come back again.

To start off with, do five sets, five times out and back,

and then rest and do that three times through.

If you start getting tired or you're losing your form

or you're wiggling around too much,

then you can just drop onto your knees.

And it's still challenging.

You can probably find you can go a little bit further.

So you're still working that core.

(upbeat music)

Well this exercise is again pretty much what it says

on the tin.

You're trying to make your body into a V shape.

So you're gonna lift your feet up,

and you tilt your upper body back slightly.

But you wanna try and still keep nice and tight here

and not cheat with your shoulders.

So I don't wanna see any round like this.

Keeping your shoulders nice and open.

You can use your hands just to help counterbalance.

Think you're trying to make a nice clean V shape.

(upbeat music)

for this exercise you can use a kettlebell,

a medicine ball, or even just something heavy

that you can hold with both your hands,

and it's all about rotation,

so strengthening the full part of your core.

So from here pick up whatever it is you're gonna use,

in my case a kettlebell.

Roll back, and you're basically trying to be

in a slightly crunched position.

So you see my head and shoulders off the ground,

my feet are off the ground.

From here, I'm gonna rotate, touch the ground on my left,

lift it up, rotate to the other side.

I want you to do 10, and by 10 I mean

this is one, two,

three, so 10

or five if you just count on this side.

Have a rest and repeat it three times.

Core exercises are often the silent part

of triathlon training.

You don't necessarily have to get hot and sweaty,

but having a strong core not only makes you

more efficient in all three sports,

it'll also help prevent you getting injured.

If you've enjoyed this, give it a thumbs up like,

and to subscribe to GTN,

just click on the globe.

If you want another body maintenance type video,

then why not watch this on the TheraBands,

or if you want a trail running video,

just click down here.

For more infomation >> 5 Core Exercises To Make Yourself Stronger | Strength And Conditioning For Triathletes - Duration: 4:35.

-------------------------------------------

Uomini e Donne trono gay, Alex e Alessandro nella bufera | Wind Zuiden - Duration: 3:41.

For more infomation >> Uomini e Donne trono gay, Alex e Alessandro nella bufera | Wind Zuiden - Duration: 3:41.

-------------------------------------------

Velshi & Ruhle December 29, 2017| MSNBC News 12/29/17 - Duration: 40:50.

For more infomation >> Velshi & Ruhle December 29, 2017| MSNBC News 12/29/17 - Duration: 40:50.

-------------------------------------------

I meravigliosi benefici dell'olio d'oliva con succo di limone - Duration: 7:35.

For more infomation >> I meravigliosi benefici dell'olio d'oliva con succo di limone - Duration: 7:35.

-------------------------------------------

GF VIP: Luca e Ivana si sono fidanzati? Ecco cos'è successo dopo il reality | WInd Zuiden - Duration: 3:33.

For more infomation >> GF VIP: Luca e Ivana si sono fidanzati? Ecco cos'è successo dopo il reality | WInd Zuiden - Duration: 3:33.

-------------------------------------------

Beauty Edu: Ceramides In Skincare - Duration: 5:16.

Hi everybody, so today's video come by way of Facebook.

I got a question from Christopher Ok and I'm like, okay Christopher!

That is the BEST NAME EVER!!!

I wish my name was Trina Alright!

Woah, er, I don't know where that came from.

Anyway!

He asked a question about tallow CLA and ceramides. Because I didn't know a lot about ceramides

and I wanted to learn more, I said…

I will make that a topic of a video.

What is a ceramide?

They are a family!

Awwwh!

A family of waxy lipid molecules.

Lipids are fats and oils that are found in the body.

They look a little something like this…

They consist of a fatty acid linked to a sphingoid base.

I LOVE that word!

And they are connected by an amide bond.

There are at least nine major ceramides present within the stratum corneum.

That is the outer most layer of skin.

They have conveniently named them ceramides 1-9.

Imagine if you had a family and you just named your kids one through nine.

This reminds me of when I was a kid and my dad wanted to name my twin sisters Alpha

and Beta.

My mom was like, "Yeah, NO."

These ceramides differ by their head group, the length of their chain, by the hydroxylation,

which is the amount of OH groups they have, and whether they're saturated or not, so that

is the number of double bonds.

You're probably wondering, where can I find these ceramides.

I was certainly wondering that.

You can find ceramides in the stratum corneum.

Inside there are these layers of dead skin cells, and sandwiched in between them is this

glue, and this glue holds the skin cells together.

It is called the interstitial lipid matrix or intercellular cement.

These lipids form these ordered structures called lamellar sheets and they alternate

between water and lipids.

And that lipid layer is composed of ceramides, cholesterols, and fatty acids.

These waxy lipids help us regulate our water holding capacity, allowing a barrier that

doesn't easily let fluid pass through it….

and this is important because we want to maintain optimal water levels in our skin.

We also want to make sure we keep disease causing microorganisms out, or for example

if we come in contact with harmful chemicals, we dont want to get them into our body.

Ceramide levels increase as we get older, so as we age we start to struggle with dry

skin.

The outer layer of the stratum corneum starts to decrease, it gets a lot thinner and we

start to notice fine lines.

Also, people who struggle with skin disorders like atopic dermatitis, eczema, and psoriasis.

Those folks seem to have a decreased level of ceramides present in their skin.

The idea behind ceramides in skincare, is that if we're missing ceramides in our skin,

maybe we can just replenish them, by adding them into products that we use.

And so cosmetic companies have taken this idea and they've run with it.

They've put ceramides into hundreds of skincare products, that make claims like increases

skin hydration, improves the cutaneous barrier, prevents moisture loss, and reduces dry and

flaking skin.

In our body their produced in the stratum granulosum.

That is below the stratum corneum.

They originally appear as phospholipids, and as they work there way up to the stratum corneum,

There are these enzymes in the stratum corneum that convert them into ceramides.

But nobody is harvesting human bodies to get these ceramides.

Instead we are getting them from other natural and synthetic sources.

Synthetics obviously come from a lab.

For natural sources, They either come from animal sources such as cows, but they are

mostly extracted from plant sources like sweet potatoes, wheat, and brown rice.

What I also learned is that ceramides are expensive, which is why we often seem them

in high end skincare.

Natural ceramides are very unstable substances, and because they are costly to obtain, synthetic

ceramides are frequently used.

Unfortunately synthetic ceramides do not penetrate the skin as well, so companies are looking

into technologies like liposomes to help them penetrate the skin.

But I don't know, what do you guys think?

Do you think it works?

Do you think it doesn't?

Tell me in the comments!

From the studies I've seen, so far, it looks like ceramides do moisturize, but I haven't

seen any evidence that they moisturize better than some of the inexpensive ingredients that

are already out on the market.

In some of these studies they're comparing ceramides to untreated skin -- and that hardly

seems fair.

They're also using higher concentrations of ceramides, then you'd expect to see

In a normal cosmetic product.

Christopher, I hope I did an okay job with this video.

If you liked this video, give it a thumbs up,

If you want to see more like it, consider subscribing, and lastly, don't forget to

click that notification bell, especially if you want to leave a comment right when that

video drops.

I will see you soon.

Have a great day.

For more infomation >> Beauty Edu: Ceramides In Skincare - Duration: 5:16.

-------------------------------------------

France Gall : hospi­ta­li­sée d'ur­gence, la chan­teuse a été admise en soins inten­sifs - Duration: 2:31.

For more infomation >> France Gall : hospi­ta­li­sée d'ur­gence, la chan­teuse a été admise en soins inten­sifs - Duration: 2:31.

-------------------------------------------

How to Polish Shoes - Care Tips That Will Make Your Shoes Happy - Duration: 4:22.

For more infomation >> How to Polish Shoes - Care Tips That Will Make Your Shoes Happy - Duration: 4:22.

-------------------------------------------

Costanza Caracciolo ha accolto il fidanzato Bobo Vieri nella sua Sicilia. La coppia ha festeggiato - Duration: 2:29.

For more infomation >> Costanza Caracciolo ha accolto il fidanzato Bobo Vieri nella sua Sicilia. La coppia ha festeggiato - Duration: 2:29.

-------------------------------------------

Smart Holiday Light Display

For more infomation >> Smart Holiday Light Display

-------------------------------------------

Lexus IS 300h F Sport Line Full Map Navigatie, Leder, Parkeersensoren - Duration: 0:54.

For more infomation >> Lexus IS 300h F Sport Line Full Map Navigatie, Leder, Parkeersensoren - Duration: 0:54.

-------------------------------------------

Sony a7R III Tested - Shooting a Mini Documentary - Duration: 4:57.

"Kazari-Kanzashi" is the traditional hair accessory made from metal

I'm making it both, for traditionally styled hair and general hair styles for women.

i'm in charge since about 25 years

and I'm working as the fourth generation in this job.

I'm successful at work

but I had no idea how to succeed at it at first.

I had another job before

but after a while I realised that this job is very important

so I decided to follow my father's footsteps.

This work should be learned by hand

rather than talked about.

I stole my father's technique and

created my own technique.

My "Kanzashi" is for traditional Japanese "Kabuki" or for "Nihon Buyou".

Since this is Japanese culture,

and it's my job to support it,

I'm making as much effort as possible to make beautiful products.

As the motif of Kanzashi,

animals or plants are often used

yes, animals or plants.

I'm watching these things every day

or touching them sometimes

and consider how to use them for Kanzashi.

A lot of customers are

women wearing a kimono.

This makes half of my customers.

And the other customers are using them on stage.

So the person named "Tokoyama"

who arranges the traditional Japanese women hair styles "Nihon-Gami".

They are my major customers.

For more infomation >> Sony a7R III Tested - Shooting a Mini Documentary - Duration: 4:57.

-------------------------------------------

19歲女星鄧月平拍戲綁鞭炮接客還慘！被吊大半天拍戲，淋到全身濕透！！ - Duration: 3:41.

For more infomation >> 19歲女星鄧月平拍戲綁鞭炮接客還慘！被吊大半天拍戲，淋到全身濕透！！ - Duration: 3:41.

-------------------------------------------

Vampire hunting - Zander fishing with Frans & Sean (English SUBS) - Duration: 7:07.

Sean: Look at that water level Frans...

Frans: Super high!

Frans: There has been a ton of rain and snow the past few weeks

We've tried to fish the river but it is super murky

You can fish there, we got some takes

But we decided to go for the lakes connected to the river

Today we are vertical jigging, targeting zander

Now and then we catch some perch as well

But our main target for today is zander

A Perca Fluviatilis!

Get over here!

Sean: We missed like 10 takes in a row

This one finally reached the boat

This one being the smallest because Frans lost a big one a few minutes ago

For real, a much bigger fish

Frans: Every fish counts!

Frans: Someone lost a softbait...

Sean: The hook was fitted by a professional 😂

Frans: Let's zoom in a little

A piece of art for sure 😁

Another zander that licked the tail of my softbait

Sean: Professionally unhooked

GOT HIM!

And its gone...

Sean: How many fish came off in a row?

Frans: 6 or 7?

Sean: How many takes in total versus the number of fish caught? 😂

Frans: I don't wanna know...

15, maybe 16...

Frustrating, we caught 1 perch and 2 zander

I think together we had like

20-25 takes during the day

It drives us crazy

****** zander 😂

Today we are fishing for zander with reasonably light gear

Well, for pike fishing standards it is light

I've got a rod from Okuma, 15-45gr

Small softbaits

0.06mm braided line

This particular softbait is a Bleak Paddle Tail from Savage Gear

In this case a 28gr jighead

Its a bit heavier than most people would use but its good to keep up some speed

while keeping it close to the bottom

You can fish with less heavier jigheads, we fished with 21gr most of the day

Small stinger close to the tail

Important since most takes where on the stinger

Not a single fish on the single hook

One of the reasons why we missed and lost many fish, they are a bit cautious

Thats it, not too complicated

We don't fish deeper than 12 meters today, which is about the maximum depth we fish

Most fish we caught today around 8 meters

Fishing between 8 and 10 meters was most effective

We just had some bad luck with the amount of fish caught

Super fun though, I think we had over 30 takes now

We got 1 hour left, let's make the most of it

I was grabbing a new bait with my rod in the rod holder

Suddenly I hear Frans yelling "Strike, strike!"

Cool fish, amazing how doing nothing can be effective

Frans: Haha again out of camera range

Sean: This really is my lucky day!

Frans: One of the benefits of that high waterlevel

Small rattling wobbler

Another attempt to video the release

Frans: Yeah, bullseye!

Sean: We are done for today!

10 fish

10 zander

and a perch

and 946 fish missed... Until next time!

For more infomation >> Vampire hunting - Zander fishing with Frans & Sean (English SUBS) - Duration: 7:07.

-------------------------------------------

PURITANIA - drum cover - Duration: 2:51.

Info about my gear in the description.