How to pre-train a deep neural network (or RNN) with unlabeled data?

Recently, I was asked about how to pre-train a deep neural network with unlabeled data, meaning, instead of initializing the model weight with small random numbers, we set initial weight from a pretrained model (with unlabeled data).

Well, intuitively, I kinda get it, it probably helps with the vanishing gradient issue and shorten the training time when there are not too much labeled data available. But still, I don’t really know how it is done, how can you train a neural network with unlabeled data? Is it something like SOM or Boltzmann machine?

Has anybody heard about this? If yes, can you provide some links to sources or papers. I am curious. Greatly appreciate!

Synchronizing actions with animations across network

I’m working on a project which allows players to take actions in time-limited turns. All players are connected to a central server, which is authoritative. That means the server sends info that a turn for given player has started and the time remaining to take an action. This leads to problems when client-side animations are involved. Each action has an associated animation, which takes some time to complete. The player should be able to initiate the action within given time, and the next player show start his turn when any animations complete. Obviously, the server has no notion of animations, but still needs to take them into account and start the next turn at proper time point. I’m not sure how to solve this problem. Possible solutions I came up with:

Constant animation time and built-in delay

The server will have a built-in delay after an action has been taken, before starting the next turn. Problems:

  • seems an artificial construct
  • animation times need to be exact

Not waiting for animations

The server can simply broadcast action effects without any delay, and go to the next turn. Problems:

  • will look very strange for the players since the effects will become visible before the animation completes
  • possible overlapping animations across turns

Sending actions along with animation times

The client can send the animation time for given action, which the server will use as a delay for the next turn. Problems:

  • obvious cheating possibility (players can use the delay for “extra time”)

So, given above, all my ideas are simply bad. Therefore the question remains: is there an existing solution for such class of problems?

Which feature/s will avoids SPAM or massive invalid transactions in a IOTA network?

Watching IOTA presentation there is a thing that is still unclear for me:

Knowing that IOTA involves PoW but a much lighter one (in order to achieve thousands of transactions per second, but also would not involve a big computational power) and also being every node that includes new transactions into a validator:

  • What will avoid to a spamer/hacker post invalid transactions into the IOTA network? Let´s say, hundreds or thousand of devices starts to include new invalid TXs (transference of tokens or data) that they got confirmed by the other nodes participating in that SPAM process.

This question is based in a IOTA presentation where a demo with a car was showed –>

I was wondering: what if a malware is installed or start to get involved into the car transactions, providing information that is not right? Knowing that the nodes that has the car are also TX “validators”, could not be possible that they generate such amount of new invalid transactions, validated by itself?

My Ubuntu 18.04 is unable to detect Wired network

I have sideloaded Ubuntu 18.04 with Windows 10 on my Surface book. When I boot Windows, it does detect the Wired network just fine, but not the same case with Ubuntu.

Could someone please help me investigate how to solve this problem?


ifconfig -a

docker0: flags=4099  mtu 1500
    inet  netmask  broadcast
    ether 02:42:30:d8:dc:e7  txqueuelen 0  (Ethernet)
    RX packets 0  bytes 0 (0.0 B)
    RX errors 0  dropped 0  overruns 0  frame 0
    TX packets 0  bytes 0 (0.0 B)
    TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

lo: flags=73  mtu 65536
    inet  netmask
    inet6 ::1  prefixlen 128  scopeid 0x10
    loop  txqueuelen 1000  (Local Loopback)
    RX packets 2527  bytes 185456 (185.4 KB)
    RX errors 0  dropped 0  overruns 0  frame 0
    TX packets 2527  bytes 185456 (185.4 KB)
    TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

wlp3s0: flags=4163  mtu 1500
    inet  netmask  broadcast
    inet6 fe80::95a:7402:dd65:5960  prefixlen 64  scopeid 0x20
    ether bc:83:85:cb:b9:1b  txqueuelen 1000  (Ethernet)
    RX packets 93574  bytes 122521784 (122.5 MB)
    RX errors 0  dropped 441  overruns 0  frame 0
    TX packets 45205  bytes 4840762 (4.8 MB)
    TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

sudo lshw -C network -numeric

   description: Wireless interface
   product: 88W8897 [AVASTAR] 802.11ac Wireless [11AB:2B38]
   vendor: Marvell Technology Group Ltd. [11AB]
   physical id: 0
   bus info: pci@0000:03:00.0
   logical name: wlp3s0
   version: 00
   serial: bc:83:85:cb:b9:1b
   width: 64 bits
   clock: 33MHz
   capabilities: pm msi pciexpress bus_master cap_list ethernet physical wireless
   configuration: broadcast=yes driver=mwifiex_pcie ip= latency=0 multicast=yes wireless=IEEE 802.11
   resources: irq:131 memory:b9500000-b95fffff memory:b9400000-b94fffff
   description: Ethernet interface
   physical id: 1
   logical name: docker0
   serial: 02:42:30:d8:dc:e7
   capabilities: ethernet physical
   configuration: broadcast=yes driver=bridge driverversion=2.3 firmware=N/A ip= link=no multicast=yes

lspci -nn | grep -i net

03:00.0 Ethernet controller [0200]: Marvell Technology Group Ltd. 88W8897 [AVASTAR] 802.11ac Wireless [11ab:2b38]

Can connect to cafe network via LAN, but cannot connect to same network via wifi [on hold]

Can you help me figure out how to get connected to a password protected network via wifi?

I can only currently access this network via the LAN.

When I try to connect via wifi, Windows tells me I cannot connect to this network.

Not sure why….

Thank you.