Hi,
>I typically see final values on the order of 0.002-0.004, I’m not
> sure whether that is normal/expected.
This is normal.
> The last column shows so far 0.0057869.... in all iterations.
All parameters tried by the Nelder–Mead algorithm are printed.
The last column does not necessarily increase in every step.
Best regards,
Takanori Nakane
On 2019/01/28 15:43, Dieter Blaas wrote:
> Hi Oli,
>
> thanks for the comment! Polishing is still running (currently at
> iteration 5). The last column shows so far 0.0057869.... in all
> iterations. I noticed in a previous run that this changes towards the
> final iterations. Yes, it would be good to know what a 'good value' is.
> Anybody could let us know?
>
> bw Dieter
>
> ------------------------------------------------------------------------
> Dieter Blaas,
> Max F. Perutz Laboratories
> Medical University of Vienna,
> Inst. Med. Biochem., Vienna Biocenter (VBC),
> Dr. Bohr Gasse 9/3,
> A-1030 Vienna, Austria,
> Tel: 0043 1 4277 61630,
> Fax: 0043 1 4277 9616,
> e-mail: [log in to unmask]
> ------------------------------------------------------------------------
>
> Am 28.01.2019 um 16:35 schrieb Oliver Clarke:
>> To be clear I’ve had excellent results with polishing too - it is only
>> the training I’ve found troublesome. In my hands the parameter choice
>> does make a difference, though it is pretty small - e.g. in one case
>> switching from s/d/a of 0.5/12000/1 improved res by 0.05 Å (from 2.8
>> to 2.75).
>>
>> I would like to know how to interpret the output of the training step
>> though - it’s not clear to me what a “good value” for the FSC column
>> is. I typically see final values on the order of 0.002-0.004, I’m not
>> sure whether that is normal/expected.
>>
>> Cheers
>> Oli
>>
>>> On Jan 27, 2019, at 3:04 PM, Dieter Blaas
>>> <[log in to unmask]> wrote:
>>>
>>> Hi Oli,
>>>
>>> thanks for the hint! Nevertheless, I had very good results (about
>>> 0.5 A better resolution) with my previous data set. I'll give it some
>>> more hours. If it does not finish, I'll try your parameters!
>>>
>>> Thanks, bw Dieter
>>>
>>> ------------------------------------------------------------------------
>>> Dieter Blaas,
>>> Max F. Perutz Laboratories
>>> Medical University of Vienna,
>>> Inst. Med. Biochem., Vienna Biocenter (VBC),
>>> Dr. Bohr Gasse 9/3,
>>> A-1030 Vienna, Austria,
>>> Tel: 0043 1 4277 61630,
>>> Fax: 0043 1 4277 9616,
>>> e-mail: [log in to unmask]
>>> ------------------------------------------------------------------------
>>>
>>> Am 27.01.2019 um 20:36 schrieb Oliver Clarke:
>>>> Hi Dieter,
>>>>
>>>> The training takes forever, sometimes days, and often converges on
>>>> nonsense results (e.g. one parameter goes to zero).
>>>>
>>>> In the past, Takanori Nakane has suggested just using the default
>>>> params. I did a couple of runs of training on particles of different
>>>> sizes on gold grids, and I usually just use those parameters with
>>>> decent results. I am sure training does add some value though…
>>>>
>>>> The parameters I use are:
>>>>
>>>> For large (MDa) particles (v/d/a): 0.5/12000/1
>>>>
>>>> For small (~150kDa) particles (v/d/a): 1/7000/0.5
>>>>
>>>> Cheers
>>>> Oli
>>>>
>>>>
>>>>> On Jan 27, 2019, at 2:30 PM, Dieter Blaas
>>>>> <[log in to unmask]> wrote:
>>>>>
>>>>> Dear all,
>>>>>
>>>>> I started the first run (i.e. training) of Bayesian Polishing
>>>>> (relion-3.1-beta) on 10,000 particles (450 x 450), 3600 micrographs
>>>>> on a machine with 8 CPUs, 128 GB RAM about 10 hours ago and it is
>>>>> still running. Sadly, there is no progress indicator as in other
>>>>> relion tasks. How long such a run will be? How does it scale? I had
>>>>> 5,000 particles previously and it finished after about 4 hours;
>>>>> output is currently showing "micrograph 482/482: 27 particles
>>>>> [10005 total]". Apart from some disk access by another program from
>>>>> another machine via sshfs, there is nothing going on on this work
>>>>> station!
>>>>>
>>>>> Thanks for hints, bw Dieter
>>>>>
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>> Dieter Blaas,
>>>>> Max F. Perutz Laboratories
>>>>> Medical University of Vienna,
>>>>> Inst. Med. Biochem., Vienna Biocenter (VBC),
>>>>> Dr. Bohr Gasse 9/3,
>>>>> A-1030 Vienna, Austria,
>>>>> Tel: 0043 1 4277 61630,
>>>>> Fax: 0043 1 4277 9616,
>>>>> e-mail: [log in to unmask]
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>>
>>>>> ########################################################################
>>>>>
>>>>>
>>>>> To unsubscribe from the CCPEM list, click the following link:
>>>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>
> ########################################################################
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|
