LETTER Communicated by Andreas Andreou Energy-Efcient Coding with Discrete Stochastic Event

LETTER
Communicated by Andreas Andreou
Energy-Ef cient Coding with Discrete Stochastic Events
Susanne Schreiber susanne@https://www.wendangku.net/doc/0a18656157.html, Institute of Biology, Humboldt-University Berlin, 10115 Berlin, Germany, and Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, U.K. Christian K. Machens c.machens@itb.biologie.hu-berlin.de Andreas V. M. Herz a.herz@biologie.hu-berlin.de Institute of Biology, Humboldt-University Berlin, 10115 Berlin, Germany Simon B. Laughlin https://www.wendangku.net/doc/0a18656157.html,ughlin@https://www.wendangku.net/doc/0a18656157.html, Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, U.K.
We investigate the energy ef ciency of signaling mechanisms that transfer information by means of discrete stochastic events, such as the opening or closing of an ion channel. Using a simple model for the generation of graded electrical signals by sodium and potassium channels, we nd optimum numbers of channels that maximize energy ef ciency. The optima depen d on several factors: the relative magnitudes of the signaling cost (current ow through channels), the xed cost of maintaining the system, the reliability of the input, additional sources of noise, and the relative costs of upstream and downstream mechanisms. We also analyze how the statistics of input signals in uence energy ef ciency. We nd that energy-ef cient signal ensembles favor a bimodal distribution of channel activations and contain only a very small fraction of large inputs when energy is scarce. We conclude that when energy use is a signi cant constraint, trade-offs between information transfer and energy can strongly in uence the number of signaling molecules and synapses used by neurons and the manner in which these mechanisms represent information.
1 Introduction
Energy and information are intimately related in all forms of signaling. Cellular signaling involves local movements of ions and molecules, shifts in their concentration, and changes in molecular conformation, all of which
c Neural Computation 14, 1323–1346 (2002) ° 2002 Massachusetts Institute of Technology

1324
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
require energy. Nervous systems have highly evolved cell signaling mechanisms to gather, process, and transmit information, and the quantities of energy consumed by neural signaling can be signi cant. In the blow y retina, the transmission of a single bit of information across one chemical synapse requires the hydrolysis of more than 100, 000 ATP molecules (Laughlin, de Ruyter van Steveninck, & Anderson, 1998). The adult human brain accounts for approximately 20% of resting energy consumption (Rolfe & Brown, 1997). Recent calculations suggest that the high rate of energy consumption in cortical gray matter results mainly from the transmission of electrical signals along axons and across synapses (Attwell & Laughlin, 2001). Given that high levels of energy consumption constrain function, it is advantageous for nervous systems to use energy-ef cient neural mechanisms and neural codes (Levy & Baxter, 1996; Baddeley et al., 1997; Sarpeshkar, 1998; Laughlin, Anderson, O’Carroll, & de Ruyter van Steveninck, 2000; Schreiber, 2000; Balasubramanian, Kimber, & Berry, 2001; de Polavieja, in press). We set out to investigate the relationship between energy and information at the level of the discrete molecular events that generate cell signals. Ultimately, information is transmitted by the activation and deactivation of signaling molecules. These are generally single proteins or small complexes that respond to changes in electrical, chemical, or mechanical potential. Familiar neural examples are the opening of an ion channel, the binding of a ligand to a receptor, the activation of a G-protein, and vesicle exocytosis. These events involve the expenditure of energy—for example, to restore ions that ow across the membrane, restore G-proteins to the inactive (GGDP) state, and remove and recycle neurotransmitter. The ability of these events to transmit information is limited by their stochasticity (Laughlin, 1989; White, Rubinstein, & Kay, 2000). This uncertainty reduces reliability and hence the quantity of transmitted information. To increase information, one must increase the number of events used to transmit the signal; this, in turn, increases the consumption of energy. We investigate this fundamental relationship between information and energy in molecular signaling systems by developing a simple model within the context of neural information processing: a population of ion channels that responds to an input by changing their open probability. We derive the quantity of information transmitted by the population of channels and demonstrate how information varies as a function of the properties of the input and the number of channels in the population. We identify optima that maximize the ratio between transmitted information and cost. These optima depend on the input statistics and, as with spike codes (Levy & Baxter, 1996), the ratio between the costs of generating signals (the signaling cost) and the cost of constructing the system, maintaining it in a state of readiness and providing it with an input (the xed costs). The article is organized as follows. Section 2 introduces the model for a system that transmits information with discrete stochastic signaling events.

Energy-Ef cient Coding
1325
In section 3, we de ne measures of information transfer, energy consumption, and metabolic ef ciency for such a system. In section 4, we analyze the dependence of energy ef ciency on the number of stochastic units for gaussian input distributions. The in uence of additional noise sources is studied in section 5. In section 6, we look at the energy ef ciency of combinations of systems, and in section 7, we derive optimal input distributions and show how energy ef ciency depends on the number of stochastic units when the distribution of inputs is optimal. Finally, in section 8 we conclude our investigation with an extensive discussion.
2 The Model
We consider an information transmission system with input x and output k. The system has N identical, independent units that are activated and deactivated stochastically. The input x directly controls the probability that a unit is activated. The number of activated units, k, constitutes the output of the system. Because realistic physical inputs are bounded in magnitude, any given distribution of inputs can be mapped in a one-to-one fashion onto the activation probabilities of units, within the interval [0I 1]. We therefore assume, without loss of generality, that x 2 [0I 1] is equivalent to the probability of being in an activated state. Consequently, the conditional probability that a given input x activates k units is given by a binomial distribution, p(k|x) D ′ N k x (1 ? x)N?k . k (2.1)
2 The variance sk|x of this binomial probability distribution, 2 sk|x D Nx(1 ? x),
(2.2)
is a measure for the system’s transduction accuracy, de ning the magnitude 2 of the noise. Note that sk|x depends on both the number of available units N and the input x. In an equivalent interpretation, the model can also be considered as a linear input-output system, k D Nx C g (N, x), (2.3)
where Nx is the “deterministic” component of the output and g (N, x) represents the noise due to the stochasticity of the units. The noise distribution corresponds to p(k|x) shifted to have zero mean; its variance is therefore 2 given by sk|x . Thus, we see that the input x speci es a noise-free output N ￠ x, to which the noise g is added, yielding the integer-valued output k.

1326
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
A
Input x determines open probability of Na+ channels stochastic
constant proportion pK open non-stochastic
B
N=5
Na+ Channel
K+ Channel
N = 10
Figure 1: (A) Schematic view of the membrane model. The input x directly determines the probability that sodium channels are open. In contrast to the stochastic sodium channels, potassium channels are considered nonstochastic at this stage of analysis; independent of the input, a constant fraction pK is open. (B) Schematic view of two model signaling systems with N D 5 and N D 10 sodium channels, respectively. Note that the ratio of sodium to potassium channels is kept constant (here NK / N D 1).
2.1 Implementing the Model. For concreteness, we have chosen to implement the model in the context of a basic neural signaling mechanism. A membrane contains two populations of voltage-insensitive channels: one controlling inward current and the other outward current (see Figure 1). For convenience, we refer to these as sodium and potassium channels, respectively. There are N sodium channels and NK potassium channels. In our analysis, an input, x, produces a voltage response by changing the open probability of the set of sodium channels, which take the role of the stochastic signaling units. The input x could be derived from a variety of sources, both external, such as light, and internal, such as synaptically released transmitter. But regardless of its origins, the input is assumed to be unambiguously represented by the open probability of sodium channels. For simplicity, the second set of ion channels—the potassium channels—is considered to be input independent and noise free. Thus, a xed proportion of potassium channels is kept open, regardless of the size of the input. The input signal speci es the voltage output signal by directly determining the probability that sodium channels are open. Thus, a given input value x, presented in a small time interval Dt , will result in k open sodium channels. Note that because of channel noise, the number of open sodium channels k will vary with different presentations of the same input. The state of the model system, on the other hand, is given by the number of open sodium channels k and translates uniquely into a voltage output if we neglect the in uence of membrane capacitance. Conversely, if we know the

Energy-Ef cient Coding
1327
output voltage V, we can directly infer the number of open sodium channels k. Both variables are therefore equivalent from an information-theoretic point of view. By working in the channel domain (i.e., by taking the number of open sodium channels k as a measure of the system’s output), we can simplify our analysis and avoid the nonlinear relationship of conductance, current, and voltage. Note that to achieve the linear relationship between the input x and the (average) output k, the channels are not voltage sensitive. For simplicity, we do not study the effects of membrane capacitance on the time course of the voltage response, assuming that the signal variation is slow in comparison to the timescale of the effects introduced by membrane capacitance. Nor do we analyze the effects of channel kinetics. By working at a fundamental level, the mapping of the input onto discrete molecular events, we can investigate a simple model of general validity.
3 Calculating the Metabolic Ef ciency
We de ne the metabolic ef ciency of a signaling system as the ratio between the amount of transmitted information I and the metabolic energy E required. Ef ciency I / E thus can be expressed in bits per energy unit. (For other ef ciency measures, see section A.3.) Both information transfer I and metabolic energy (or cost) E depend on the number of available signaling units—the number of channels, N. To investigate the relationship between ef ciency and the number of stochastic units, we drive the model system with a xed distribution of inputs, px (x), and vary N, the size of the population of sodium channels used for signaling. This is equivalent to changing the channel density of our model membrane while maintaining the same input. To ensure that systems with different values of N, the number of sodium channels, produce the same mean voltage output in response to a given input, x, the population of potassium channels, NK , is set to a constant proportion of N. Under these conditions, we can now calculate how the information transmitted, I, and the energy consumed, E, vary for different numbers of channels.
3.1 Information Transfer. Consider the transmission of a single signal. The model system receives an input, x, drawn from the input distribution, px (x), and produces a single output, k. According to Shannon, the information transmitted by the system is given by
I[NI px (￠)] D
N XZ kD 0
1 0
dxp (k|x)px (x) log2
μ
? p(k|x) , p k (k)
(3.1)
and depends on the input distribution px (x) and, via the binomial distribution of channel activation p(k|x) (see equation 2.1), also on the number

1328
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
of available units, N. For a continuously signaling neuron, this is equivalent to the response during a discrete time bin of duration Dt , and I is the rate of transmission in bits/Dt . In this article, we study two different scenarios—gaussian input distributions and input distributions that maximize the information transfer—both in the presence of noise caused by stochastic units (e.g., channel noise).
3.2 Energy. The total amount of energy required for the maintenance of a signaling system and the transmission of a signal is given by
E[NI px (￠)] D b C
Z
1 0
px (x)e(x, N) dx,
(3.2)
where b is the xed cost, and e(x, N) characterizes the required energy as a function of the input x and the number of stochastic signaling units N. Thus, we classify the metabolic costs into two groups: costs that are independent of N and costs that depend on the total number of channels, N. For simplicity, we assume that the latter costs are dominated by the energy used to generate signals (in this case, restoring the current that ows through ion channels), and we neglect the energy required to maintain channels. The rst group of costs, b, relates to costs that have to be met in the absence of signals, such as the synthesis of proteins and lipids. These costs are therefore called xed costs and are constant with respect to x and N. Because we have set up our systems to produce identical mean values of the voltage output signal given x (by xing the ratio N / NK ), the function e(N, x) is separable into the variables N and x (see section A.1), e(x, N) D NQ (x), e (3.3)
so that the signaling-related total energy consumption rises linearly with N. Q Q Q The function e (x) increases monotonically between e (0) D 0 and e (1) D ech , where ech denotes the energy cost associated with a single open sodium Q channel. (The precise form of e (x) is derived in section A.1.) Rescaling the measurement units of energy, we will from now on set ech D 1. Altogether, the total energy reads E[NI px (￠)] D b C e (x)N, Q (3.4)
Q where e (x) is the average signal-related energy requirement of one stochastic unit and the average is taken with respect to the input distribution px (x). In the rst part of the analysis, where we analyze energy-ef cient numbers of channels, we make the simplifying assumption that the average cost per Q channel, e (x), is approximately equal to the mean of the input, x. Note that the energy E is de ned as a measure of cost for one time unit Dt , just as I is the measure of information transfer in Dt.

Energy-Ef cient Coding 4 Gaussian Input Distributions
1329
4.1 Information Transfer. We focus on gaussian inputs rst, because according to the central limit theorem, they are a reasonable description of signals composed of many independent subsignals, and they also allow an analytic expression of information transfer. To con ne the bulk of the gaussian distributions within the input interval [0I 1], the mean, x, and the 2 variance,sx , are chosen such that the distance from the mean to the interval borders is always larger than 3sx . Values falling outside the interval [0I 1] are cut off, and the distribution is then normalized to unity. Numerical simulations (see section A.2) show that the effects of this procedure on the results are negligible. The information transfer I (Shannon & Weaver, 1949) per unit time for a linear system with additive gaussian noise and gaussian inputs is given by
ID
1 log2 (1 C SNR), 2
(4.1)
where SNR denotes the signal-to-noise ratio. It is de ned as the ratio be2 2 tween the signal variance, sx , and the effective noise variance, sk|x / N 2 . If two criteria are met— rst, the binomial noise g (N, x) can be approximated by a gaussian and, second, within the regime of most likely inputs x, changes in 2 the noise variance sx| k are negligible—the following equation gives a reasonably good approximation of the information transfer of our model system: ID
2 Nsx 1 log2 1 C x(1 ? x) 2
′
.
(4.2)
This is the case for large N and a restriction to the gaussian inputs described above. Numerical tests (see section A.2) show that the deviation between the real information transfer with N stochastic signaling units and the information transfer given by equation 4.2 is very small. expression:
4.2 Ef ciency. For the ef ciency, de ned as I / E, we obtain the following
2 I Nsx 1 D log2 1 C E x(1 ? x) 2(xN C b)
′
.
(4.3)
In Figure 2, ef ciency curves I / E are depicted as a function of N for three different values of xed costs b. The energy ef ciency exhibits an optimum for all curves. Signal transmission with numbers of channels N within the range of the optimum is especially energy ef cient. The position of the optimum depends strongly on the size of the xed cost b relative to the average Q cost of a single channel e (x) ? x. Figure 2 displays the dependence of the

1330
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
optimal N
0.004
A
b=500
2000 1000 2500 5000
efficiency I/E
0.003 0.002
fixed cost b
b=2000
0.001
b=5000
0 2000 4000 6000 8000 10000
number of channels N
normalized efficiency
B
1 0.8 0.6 0.4 0.2 0 (I/E)n
1 0.5 0
b=5000
b=50
log(N/optimal N) 1
-5
0
5
N / optimal N
2
3
4
5
Figure 2: (A) Ef ciency I / E as a function of N for three different xed costs. From top to bottom, the xed cost b is equivalent to the cost of 500, 2000, and 5000 open channels. The optimal N shifts to larger N with rising xed cost b (see also the inset), and the ef ciency curves become wider. (B) Ef ciency curves for xed costs of b D 50, b D 500, and b D 5000 rescaled, so that the maximum corresponds to (1,1). The inset shows the same data on a logarithmic scale. These rescaled curves, (I / E) n , are similar in shape, independent of the size of b. For all graphs, the parameters of the input distribution px (x) are sx D 0.16 and x D 0.5.
optimal number of channels N on the xed cost b. The most ef cient number of channels increases approximately linearly with the size of the xed cost, although close inspection reveals that the slope is steeper for smaller xed costs and shallower in the range of higher b. If b D 0, the most ef cient system capable of transmitting information uses one channel. The average costs of the most energy-ef cient population of channels, employing Nopt channels, are given by Nopt x. Therefore, the ratio between these signaling N costs at x D 0.5, and the xed cost b is approximately 1:4 in the example depicted. This ratio however, which can be derived from the slope of the

Energy-Ef cient Coding
1331
b-Nopt curve in the inset of Figure 2A, strongly depends on the input distribution. The analysis for gaussian input distributions shows that the ratio of 2 Nopt to b increases with decreasing input variance, sx (results not shown). Remarkably, the ef ciency curves rise very steeply for small N and, after passing the optimum, decrease only gradually. This characteristic does not strongly depend on the size of b, as shown in Figure 2B. It is thus very uneconomical to use a number of channels suf ciently far below the optimum, whereas above, there is a broad range of high ef ciency. In this range, a cell could adjust its number of channels to the amount of information needed (e.g., add more channels), without losing much ef ciency. However, as the inset to Figure 2B indicates, increasing the number of channels by a given factor has a similar effect on ef ciency as decreasing them by the same factor.
5 Additional Noise Sources
The most energy-ef cient number of channels is in uenced by the size of additional noise. This might be noise added to the input (additive input noise) or internal noise generated within the signaling system independent of the activation of sodium channels (input-independent internal noise).
5.1 Additive Input Noise. If the input contains gaussian noise of a xed 2 variance hgx i that is not correlated with the signal, the variances of the signal and the noise add, yielding the modi ed SNR at the output
SNR D
2 Nhgx i C
2 Nsx . x(1 ? x)
(5.1)
The additional noise gx decreases the SNR and, consequently, the informa2 2 tion transfer. For N ! 1, the SNR converges to sx / hgx i, and thus sets an upper limit to the information transfer. Figure 3 shows that it is more ef cient to operate at lower numbers of channels in the presence of additional signal noise.
5.2 Input-Independent Internal Noise. To exemplify internal noise, we consider an additional population of sodium channels that is not in uenced by the input x but rather has a xed open probability p0 , though it contributes to the noise. Assuming a xed ratio N / N 0 between the total number of the original input-dependent sodium channels N and the total number of these new input-independent sodium channels N 0 , the voltage output is determined by the sum of open channels from both populations k C k0 . Because noise from both populations is uncorrelated, the SNR reads
SNR D
2 N 2 sx , N 2 hg2 i C Nx(1 ? x) C N0 p0 (1 ? p0 ) x
(5.2)

1332
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
efficiency I/E
0.0006 0.0004 0.0002
noiseless input
noisy input
0
2000
4000
6000
8000
10000
number of channels N
Figure 3: Ef ciency I / E as function of the number of signaling units, N, for the case of a noisy input signal (solid line). For comparison, we also replot the case where there is no input noise (dotted line), as in Figure 2A. For both curves, the variance of the input distribution equals sx2 D 0.162 , x D 0.5, and b D 5000. The 2 noise variance of the signal is hgx i D 0.042 .
where N 0 p0 (1 ? p0 ) represents the noise variance of the input-independent population of channels. The ef ciency of the signaling system is thus further decreased for all N, and the ef ciency optimum, Nopt , is shifted to lower values. Other noise sources, such as leak channels and additional synaptic inputs, will also lower the SNR. Therefore, they will reduce ef ciency and in uence the optimum number of channels.
6 Ef ciency in Systems Combining Several Mechanisms
Signaling mechanisms do not act in isolation; they are usually organized into systems in which one mechanism drives another, either within a cell or between cells. The relationship between information transfer and cost of each mechanism determines optimal patterns of investment in signaling units across the whole system, as we will demonstrate with some simple examples. First, consider two signaling mechanisms in series (see Figure 4). Cell 1 uses N1 channels to convert the input x into the output k1 , which, in turn, drives cell 2, which uses N2 channels, to produce the output k2 . From equation 2.3, we know that k1 D N1 x C g1 , where Nx is the signal and g1 is the noise generated by the random activity of channels. Because we de ne an input in terms of a probability distribution of signals, ranging from 0 to 1, the output k1 of cell 1 should be normalized by N1 , so that the input to cell 2 is k1 / N1 . Note that, for simplicity, we are neglecting nonlinearities in signal transfer within a cell, as, for example, in neurotransmitter release. As a consequence, the mean open probability in both cells is the same, but its variance differs. The output of cell 2 is given by k2 D N2 x2 C g2 , where g2

Energy-Ef cient Coding
1333
channel noise
1
channel noise
2
output k 1
Cell 1 Cell 2
output k 2
...
input x
input x 2 x 2 = k1 / N1
input x 3 x 3 = k2 / N2
Figure 4: Schematic view of the cell arrangement . The normalized output of cell 1 serves as input for cell 2. Both cells are subject to channel noise g1 and g2 , respectively.
is the additive channel noise of cell 2. Therefore, the information transfer 2 from an input signal x, with mean x and variance sx , to the output k2 can be approximated by Shannon’s formula as ID
2 N1 N2 sx 1 log2 1 C (N1 C N2 )x(1 ? x) 2
′
.
(6.1)
The cost of information transfer through the two-cell system is E D ech1 xN 1 C ech2 xN2 C b1 C b2 , (6.2)
where b1 and b2 are the xed metabolic costs of the cells and ech1/ 2 x the costs per open channel. i If we introduce an effective number of channels, Neff D N1 N2 / (N1 C
E N2 ), for the information transfer and Neff D N1 C N2 for the metabolic cost, the equations for I and E correspond to those of the single-cell case— equations 4.2, and 3.4, respectively. For simplicity, the cost per open channel E i is set to unity for both cells. Because Neff ? Neff for all nonnegative N1 and N2 , the information transfer increases more slowly with the number of channels than the cost, cutting down ef ciency. Thus, a two-cell system requires more channels to transmit the same amount of information and is therefore less ef cient than a single cell, even if the xed cost of a cell in the two-cell model is only half the cost of the single cell. Consequently, in a metabolically ef cient nervous system, serial connections should be made only when signal processing is required.

1334
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
Furthermore, in an energy-ef cient system, the cost of one mechanism in uences the amount of energy that should be consumed by another. Such considerations are important when one set of signaling events is more costly than another. This can be demonstrated by incorporating the cost of an upstream mechanism into the xed cost of a downstream mechanism (i.e., the cost of a mechanism includes the cost of providing it with a signal). Here we can de ne E as E D ech2 xN2 C b¤ 2 with b¤ D ech1 xN 1 C b1 C b2 . 2 (6.3)
As we have seen with one mechanism, an increase in xed cost raises the optimal number of channels. Therefore, when the cost of an upstream mechanism is high, one improves energy ef ciency by using more units downstream. The precise pattern of investment required for optimal performance will depend on the relative costs ( xed and signaling) of every mechanism (e.g., channels) and the characteristics of the input signal.
7 Limits to the Achievable Ef ciency
Information transfer and ef ciency depend on the distribution of input signals. In the previous sections, we have considered gaussian inputs. We now calculate the input distribution that maximizes information transfer I given a limited energy budget E and a particular number of channels N. The ef ciency I / E reached gives an upper bound on the ef ciency the system can achieve for given E and N. Although the nervous system has less in uence on the distribution of external signals, it is able to shape its internal signals to optimize information transfer (Laughlin, 1981). opt The optimal input distribution, px (x), and the maximum information transfer (the information capacity CN ) of a system with N stochastic units can be obtained by the Blahut-Arimoto algorithm (Arimoto, 1972; Blahut, 1972), which is described in further detail in section A.4. Given the noise distribution p(k|x), the algorithm yields a numerical solution to the optimization of the input distribution px (x), maximizing the information transfer and minimizing the metabolic cost. This algorithm has been applied by Balasubramanian et al. (2001) and de Polavieja (in press) to study the metabolic ef ciency of spike coding.
7.1 Optimal Input Distributions. For a given number of channels N, a Q given xed cost b, and a given cost function depending on the input e (x), Q the energy E D N e (x) C b used by the system depends exclusively on the input distribution px (x). If the available energy budget is suf ciently large, energy constraints do not in uence the shape of the optimal input distribution, and the information capacity CN of the system reaches its maximum (see Figure 5E, point A). The optimum input distribution turns out to be symmetrical, with inputs from the midregion around x D 0.5 drawn less of-

Energy-Ef cient Coding
0.4 0.4
1335
probability
0.3 0.2 0.1
0.2 0.4 0.6 0.8 1
input x
1 1
probability
A
0.3 0.2 0.1
B
0.2 0.4 0.6 0.8
input x
1
probability
0.8 0.6 0.4 0.2
0.2 0.4 0.6 0.8
input x
1
probability
C
0.8 0.6 0.4 0.2
D
0.2 0.4 0.6 0.8
input x
1
information capacity CN
4
E
C B A
3
2
1
D
0
10
20
energy E
30
40
50
60
Figure 5: (A)–(D) Optimal input distributions for different energy budgets E with b D 0 and N D 100. The distributions were calculated numerically and are discretized to a resolution Dx D 0.01. All distributions show small values for 2 inputs around x D 0.5 where the noise variance sk|x is highest. With decreasing energy budget, the distributions become less symmetrical, preferring low inputs. (E) Information capacity CN D 100 depending on the energy budget (for b D 0). The points mark the location of the input distributions shown in A–D in the energy capacity space.
ten, whereas inputs at the boundaries 0 and 1 are preferred (see Figure 5A). This result is very intuitive, when we take the dependence of the noise vari2 ance, sk|x , as de ned in equation 2.2, on the input x into account. The noise variance is symmetrical as well, showing a maximum at x D 0.5 and falling off toward x D 0 and x D 1. Thus, an input distribution that is optimal from the point of view of information transfer, leaving metabolic considerations aside for a moment, favors less noisy input values over noisier ones.

1336
5 4 3 2 1 0 b = 20
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
optimal N
efficiency I/E
A
0.02
B
Optimal Gaussian
500 250
fixed cost b
500
1000
0.01
energy
40
60
80
0 0
number of channels N
500
1000
Figure 6: (A) Capacity CN D 100 (of Figure 5E) as a function of energy E for a xed cost b D 20 (solid line). The capacity curve is simply shifted along the energy axis by the value of b. The ef ciency CN / E as a function of E is shown as a dash-dotted line. The maximum ef ciency (CN / E) max is given at the point of the CN (E) curve whose tangent (dashed line) intersects with the origin. (B) Achievable ef ciency (CN / E) max as a function of N for a xed cost b D 200 (solid line). For comparison, the ef ciency I / E for a gaussian input distribution (N D 0.5, sx D 0.16) is also x shown (dashed line). The inset depicts the optimal number of channels Nopt as a function of b.
Limiting the energy that can be used by a system of N units, however, changes the optimal input distribution, px (x), and destroys the symmetry. As we reduce the energy budget, values neighboring x D 0 are increasingly preferred and the costly values approaching x D 1 are avoided (see Figures 5B–5D). This asymmetry reduces the information capacity CN . Thus, ef cient use of a restricted budget requires a cell to keep most of its units deactivated. For our simple model, this is equivalent to maintaining the membrane close to resting potential by keeping most of its sodium channels closed. The xed cost, b, is a metabolic cost independent of the value of the input x and cannot be avoided. Consequently, it does not in uence the shape of the energy-capacity curve of a system. Adding the xed cost b results merely in a horizontal translation of the energy-capacity curve (see Figure 6A). Here as well, the shape of the input distribution changes with the value of E.
7.2 Ef ciency. Having obtained the dependence of the information capacity on the energy used for a given N, we can also derive energy ef ciency CN / E as a function of the energy E. Of particular interest is the maximum value (CN / E) max D maxE fCN / Eg, giving the optimal ef ciency for a xed value of N, which is achieved by a speci c input distribution px (x). Note that this ef ciency gives the upper bound to the achievable ef ciency in our system and therefore cannot be surpassed by any other input distribution px (x).

Energy-Ef cient Coding
1337
At the maximum of CN / E, the rst derivative with respect to E is zero,
@ @E
CN E
′
D 0,
(7.1)
which can be transformed to give
@CN @E
￠ E D CN .
(7.2)
Thus, geometrically, the maximum value (CN / E)max corresponds to the point on the capacity graph whose tangent intersects the origin (see Figure 6A). The slope of the tangent is given by (CN / E)max itself, so that the optimal ef ciency (CN / E)max decreases with increasing b, as can be inferred from Figure 6A by shifting the CN (E) curve to the right. Figure 6B shows the optimal ef ciency (CN / E)max as a function of the number of channels N for a xed cost b D 200. For comparison, we also show the ef ciency I / E obtained for a gaussian input distribution. The optimal input distribution surpasses the gaussian distribution roughly by a factor of two in this case. In conclusion, as with gaussian input distributions, we can derive the most ef cient number of channels Nopt as a function of the xed cost. The use of optimal input distributions reduces Nopt but, as with gaussian inputs, Nopt rises approximately linearly with the basic cost, b, as illustrated in the inset of Figure 6B.
8 Discussion
A growing number of studies of neural coding suggest that the consumption of metabolic energy constrains neural function (Laughlin, 2001). Comparative studies indicate that the mammalian brain is an expensive tissue whose evolution, development, and function have been shaped by the availability of metabolic energy (Aiello & Wheeler, 1995; Martin, 1996). The human brain accounts for about 20% of an adult’s resting metabolic rate. In children, this proportion can reach 50% and in electric sh 60% (Rolfe & Brown, 1997; Nilsson, 1996). Because much of this energy is used to generate and transmit signals (Siesjo, 1978; Ames, 2000), these levels of consumption have ¨ the potential to constrain neural computation by placing an upper bound on synaptic drive and spike rate (Attwell & Laughlin, 2001). Current evidence suggests that the advantages of an energy-ef cient nervous system are not con ned to larger animals with highly evolved brains. In general, the speci c metabolic rate of brains is 10 to 30 times the average for the whole animal, at rest (Lutz & Nilsson, 1994). In addition, for both vertebrates and insects, the metabolic demands of the brain are more acute in smaller animals because the ratio of brain mass to total body mass de-

1338
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
creases as body mass increases (Martin, 1996; Kern, 1985). Moreover, insect species with similar body masses exhibit large differences in the total mass of the brain and in the masses or relative volumes of different brain areas, and these differences correlate with behavior and ecology (Kern, 1985; Gronenberg & Liebig, 1999; Laughlin et al., 2000). Signi cant changes have been observed among individuals of a single species. In the ant Harpegnathos, the workers are usually visually guided foragers. However, when young workers are inseminated and begin to lay eggs following the death of their queen, their optic lobes are reduced by 20% (Gronenberg & Liebig, 1999). These observations suggest that the reduction of neural energy consumption is also a signi cant factor in the evolution of small brains. The relationship between energy consumption and the ability of neurons to transmit information suggests that nervous systems have evolved a number of ways of increasing energy ef ciency. These methods include redundancy reduction (Laughlin et al., 1998), the mix of analog and digital operations found in cortical neurons (Sarpeshkar, 1998), appropriate distributions of interspike intervals (Baddeley et al., 1997; Balasubramanian et al., 2001; de Polavieja, 2001), and distributed spike codes (Levy & Baxter, 1996). We have extended these previous theoretical investigations to a level that is more elementary than the analysis of signaling with spikes: the representation of information by populations of ion channels. Thus, as independently advocated by Abshire and Andreou (2001), we have analyzed the energy ef ciency of information transmission at the level of its implementation by molecular mechanisms. We have estimated the amount of information transmitted by a population of voltage-insensitive channels when their open probability is determined by a de ned input. These channels typify the general case of signaling with stochastic events. Consequently, our analysis is also applicable to many other forms of molecular signaling (e.g., the binding of a ligand to a receptor) and to synaptic transmission (e.g., the release of a synaptic vesicle according to Poisson or binomial statistics). Our theoretical results verify a well-known trend: the amount of information carried by a population of channels increases with the size of the population because random uctuations are averaged out, as observed in blow y photoreceptors and their synapses (de Ruyter van Steveninck, Lewen, Strong, Koberle, & Bialek, 1997; Laughlin et al., 1998; Abshire & Andreou, 2000) and demonstrated by models of synaptic transmission to cortical neurons (Zador, 1998). However, increasing the number of channels in the population increases both the level of redundancy and the energy used for transmission, leading to changes in metabolic ef ciency (Laughlin et al., 1998). Following Levy and Baxter (1996), we have chosen to discuss ef ciency as the ratio between the number of bits of information transmitted and the energy consumed. However, our analysis also provides a mathematical framework to describe energy ef ciency from the more general point of view of maximizing information transfer and minimizing the metabolic cost

Energy-Ef cient Coding
1339
(maximizing I ? sE, where s describes the importance of energy minimization over information maximization), as brie y outlined in section A.3. We distinguish two energy costs: the cost of generating signals and the xed cost of keeping a signaling system in a state of readiness. The signaling cost is derived from the current ow through ion channels. Under the assumptions of the model, the signaling cost increases with the number of channels. This simple linear relationship can be easily applied to other forms of signaling, such as protein phosphorylation or the turnover of transmitter and second messenger. In the absence of data on the costs of constructing and maintaining a population of channels in a membrane, we again follow Levy and Baxter (1996). The xed cost is expressed in the same units as the signaling cost and is varied to establish its effect on ef ciency. The analysis demonstrates that energy ef ciency is maximized by using a speci c number of channels. These optima depend on a number of important factors: the xed cost of the system, the cost of signaling, the reliability of the input, the amount of noise generated by other intrinsic mechanisms, the cost of upstream and downstream signaling mechanisms, and the distribution of the input signals provided by upstream mechanisms. Each of these factors is involved in neural processing. The xed cost of building and maintaining the cell in a physiological state within which the ion channels operate is a dominant factor. When the xed cost increases, the optimum system increases the energy invested in signaling by increasing the number of channels (see Figure 2A). Levy and Baxter (1996) discovered the same effect in their analysis of energy-ef cient spike trains. This may well be a general property of energy-ef cient systems because when a signaling system is expensive to make and maintain (i.e., the ratio between xed and signaling costs is high), it pays to increase the return on the xed investment by transmitting more bits. This takes more channels and more signaling energy. For the example shown, which operates with a broad input distribution and a mean open probability of 50%, the optimum population of channels has a peak energy consumption (all channels open) that is approximately half the xed cost (see Figure 2A). The relationship between the energy spent on signaling by the channels and the xed cost varies with the distribution of inputs. For input distributions that make a reasonably broad use of possible open probabilities, the ratio between signaling costs and xed costs lies approximately in the range between 1:4 and 1:1. It is dif cult to judge whether populations of neuronal ion channels follow this pattern because data about the ratio of xed costs to signaling costs in single cells are not available. However, in more complicated systems, the proportion of energy devoted to signaling is in the predicted range. For the whole mammalian brain, signaling has been linked to approximately 50% of the metabolic rate (Siesj¨ 1978), and in cortical gray matter this rises o, to 75% (Attwell & Laughlin, 2001). We are aware that there are a number of additional factors, not accounted for by our model, that will in uence the ratio of signaling costs to xed costs

1340
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
in nervous systems. For example, our analysis underestimates the total energy usage by the brain because it is con ned to a single operation: the generation of a voltage signal by channels. Within neural systems, signals must be transmitted over considerable distances, and various computations must be performed. These extra operations take extra energy. Along these lines, the transmission of signals along axons, in the form of action potentials, accounts for over 40% of the signaling cost of cortical gray matter (Attwell & Laughlin, 2001). Noise at the input reduces the optimum number of channels (see Figure 3) because it reduces the effect of channel uctuations on the total noise power. There is some evidence that nervous systems reduce the number of signaling events in response to a less reliable input. In the blow y retina, the SNR of chromatic signals is lower than that of achromatic signals, and a class of interneurons involved in the chromatic pathway uses fewer synapses than comparable achromatic interneurons (Anderson & Laughlin, 2000). Considering a chain of signaling mechanisms allows us to study networks where the output from one population of channels de nes the SNR at the input of the next. As a result, when analog signals are transferred from one mechanism to another, the noise accumulates stage by stage (Sarpeshkar, 1998). The chain describes how this buildup of noise reduces metabolic ef ciency. Given this reduction, an energy-ef cient system should connect one population of channels (or synapses) to another in a serial manner only when information is actually processed, not when it is merely transmitted. Where signals must be repeatedly ampli ed to avoid attenuation, pulses should be used to resist the buildup of analog noise (Sarpeshkar, 1998). These design strategies are hypothetical. The energy savings that are made by restricting the number of serial or convergent analog processes and converting analog signals to spikes (Sarpeshkar, 1998; Laughlin et al., 2000) have yet to be demonstrated in a neural circuit. Our analysis suggests that when transmission and processing involve several types of signaling events (e.g., calcium channels, synaptic vesicles, and ligand gated channels at a synapse), it is advantageous to use more of the less expensive events and fewer of the more expensive. This distribution is analogous to the pattern of investment in bones of a mammal’s leg. Proximal bones are thicker than distal bones because they are less costly per unit mass (they move less during a stride). The thickening of distal bones is adjusted to optimize the ratio between the probability of fracture and cost for the limb as a whole (Alexander, 1997). Finally, the probability distribution of the input signal has a large effect on ef ciency. On both an evolutionary timescale as well as on the timescale of physiological adaptation processes, the way an external signal with speci c statistical properties is transmitted could therefore be optimized by mapping the external signal distribution on an ef cient distribution of probabilities of channels to be open (which we call the input distribution). More

Energy-Ef cient Coding
1341
importantly, the internal signals passed on from one mechanism to the next could be shaped such that the signal distributions employed will enhance the ef ciency of information transfer. The Blahut-Arimoto algorithm yields input distributions that optimize the amount of information transferred under a cost constraint and has been successfully applied to spike codes (Balasubramanian et al., 2001; de Polavieja, 2001). Our application shows how inputs can be mapped onto the probabilities of activating signaling molecules to maximize the metabolic ef ciency of analog signaling. The improvement over gaussian inputs is greater than 50% and is achieved by two means. First, signaling avoids the midregion of activation probabilities where, according to binomial statistics, the noise variance is high. Second, signaling avoids the expensive situation of having a high probability of opening channels in favor of the energetically cheaper low-probability condition, similar to the results that a metabolically ef cient spike code avoids high rates (Baddeley et al., 1997; Balasubramanian et al., 2001). Our analysis suggests that an ef cient population of sodium and potassium channels usually operates close to resting potential, with most of its sodium channels closed, but infrequently switches to opening most of its sodium channels. In other words, there is a tendency toward using a combination of numerous small signals close to a low resting potential and less frequent voltage peaks. In conclusion, we have analyzed the energy ef ciency of a simple biological system that represents analog signals with stochastic signaling events, such as ion channel activation. Optimum con gurations depend on the basic physics that connects information to energy (the dependency of noise, redundancy, and cost on the number of signaling events) and basic economics (the role played by xed costs in determining optimum levels of expenditure on throughput). Given this fundamental basis, the principles that we have demonstrated are likely to apply to other systems. In particular, we have shown that energy ef ciency is a property of both the component mechanism and the system in which it operates. To assess a single population of ion channels, we had to consider the xed cost, the distribution of signal and noise in the input, and additional noise sources. After connecting two populations of ion channels in series, we had to add the relative costs of the two mechanisms and the noise fed from one population to the next to our list of factors. Energy ef ciency is achieved by matching the mechanism to its signal or, for optimum input distributions, the signal to the mechanism. Matching is a characteristic of designs that maximize the quantity of information coded by single cells, regardless of cost. To achieve this form of ef ciency, neurons exploit the adaptability of their cellular and molecular mechanisms (Laughlin, 1994). The extent to which the numbers of channels and synapses used by neurons, and their transfer functions, are regulated for metabolic ef ciency remains to be seen. The analysis presented here provides a starting point that can guide further experimental and theoretical work.

1342 Appendix
S. Schreiber, C. K. Machens, A. V. M. Herz, and S. B. Laughlin
A.1 Energetic Cost of Inputs. For the channel model, the average energetic cost per unit time e(x, N) is a function of the input x and the number of sodium channels N. The sodium and potassium currents, iNa and iK , depend on the reversal potentials ENa and EK , the membrane potential V, as well as on the conductances of the sodium and potassium channels, respectively,
iNa (x) D NgNa0x(V(x) ? ENa ),
(A.1) (A.2)
iK (x) D NK gK0 pK (V(x) ? EK ).
The vigorous electrogenic pump extrudes three sodium ions and takes up two potassium ions for every ATP molecule hydrolyzed. The pump current ipump equals iK / 2, assuming that the pump maintains the internal potassium concentration by accumulating potassium ions at a rate equal to the outward potassium current, iK . Equating all currents across the membrane gives iNa (x) C iK (x) C ipump (x) D iNa (x) C 3 iK (x) D 0. 2 (A.3)
The energetic cost e(x, N) is proportional to the pump current, ipump , so that we de ne e(x, N) D c ￠ ipump (x)
D cN ￠
(A.4) ±
NK N
gK0 pK
3gK0 pK
2
where c is the factor of proportionality. Because ( NK ) is constant, we can N separate e(x, N) into the variables N and x: e(x, N) D NQ (x). e Q The energy function e (x) can be written as Q e (x) D C ABx , Ax C B (A.7) (A.6)
±
gNa0 (ENa ? EK )x 2 , NK C 2gNa0 x N
(A.5)
with the constants A D 2gNa0, B D 3gK0 pK ( NK ), and C D c(ENa ? EK ) / 6. N Q Because we de ne units of energy in this study such that e (1) D ech D 1, the rescaled energy function that is implemented in the Blahut-Arimoto algorithm reads Q e (x) D (A C B)x . Ax C B (A.8)
It does not depend on the values of the reversal potentials ENa and EK .

软件工程毕设模板(软件开发类)v1.0

摘要 提示：摘要要点如下： 第一句：系统所依赖的背景； 第二句：设计并开发了XXX系统，主要功能是什么？ 第三句：系统实现了XXX，YYY，ZZZ的功能模块； 第四句：系统测试结果怎样？有没有符合预期？ 关键字XXX、XXX、XXX、XXX 提示：关键字应该具有代表性，建议在3-5个

目录 第1章绪论 ........................................................... 错误！未定义书签。 1.1 选题背景与意义............................................................... 错误！未定义书签。 提示：选题的背景、现状、意义 1.2 论文的主要工作............................................................... 错误！未定义书签。 提示：研究内容及章节安排 第2章相关技术和开发环境........................................... 错误！未定义书签。 2.1 相关技术......................................................................... 错误！未定义书签。 2.1.1 相关技术1 ................................................................. 错误！未定义书签。 2.1.2相关技术2 .................................................................. 错误！未定义书签。 2.1.3相关技术3 .................................................................. 错误！未定义书签。 2.1.4相关技术4 .................................................................. 错误！未定义书签。 提示：相关技术应该选取具有代表性，同时数量建议在3-5个之间 注意：该部份注意查重 2.2 开发环境........................................................................... 错误！未定义书签。 提示：开发的软、硬件环境，同时对一些关键的环境设置和开源包进行说明第3章系统分析 ................................................... 错误！未定义书签。 3.1 可行性研究....................................................................... 错误！未定义书签。 3.1.1经济可行性 ................................................................ 错误！未定义书签。 3.1.2 技术可行性 ................................................................ 错误！未定义书签。 3.1.3 运行可行性 ................................................................ 错误！未定义书签。 3.2 系统需求分析................................................................... 错误！未定义书签。 3.2.1功能需求分析 ............................................................. 错误！未定义书签。 提示：提供完整的功能需求、辅助必要的用例图 3.2.2 业务需求分析 ............................................................ 错误！未定义书签。 提示：进行业务流程分析、提供必要的流程图 3.2.3 数据需求分析 ............................................................ 错误！未定义书签。 提示：分析系统的数据需求，提供必要的数据流图 第4章概要设计 ................................................... 错误！未定义书签。

对翻译中异化法与归化法的正确认识

对翻译中异化法与归化法的正确认识 班级：外语学院、075班 学号：074050143 姓名：张学美 摘要：运用异化与归化翻译方法,不仅是为了让读者了解作品的内容,也能让读者通过阅读译作,了解另一种全新的文化,因为进行文化交流才是翻译的根本任务。从文化的角度考虑,采用异化法与归化法,不仅能使译文更加完美,更能使不懂外语的人们通过阅读译文,了解另一种文化,促进各民族人们之间的交流与理解。翻译不仅是语言符号的转换,更是跨文化的交流。有时,从语言的角度所作出的译文可能远不及从文化的角度所作出的译文完美。本文从翻译策略的角度，分别从不同时期来说明人们对异化法与归化法的认识和运用。 关键词:文学翻译；翻译策略；异化；归化；辩证统一 一直以来,无论是在我国还是在西方,直译(literal translation)与意译(liberal translation)是两种在实践中运用最多,也是被讨论研究最多的方法。1995年,美籍意大利学者劳伦斯-韦努蒂(Lawrence Venuti)提出了归化(domestication)与异化(foreignization)之说,将有关直译与意译的争辩转向了对于归化与异化的思考。归化与异化之争是直译与意译之争的延伸,是两对不能等同的概念。直译和意译主要集中于语言层面,而异化和归化则突破语言的范畴,将视野扩展到语言、文化、思维、美学等更多更广阔的领域。 一、归化翻译法 Lawrwnce Venuti对归化的定义是,遵守译入语语言文化和当前的主流价值观,对原文采用保守的同化手段,使其迎合本土的典律,出版潮流和政治潮流。采用归化方法就是尽可能不去打扰读者,而让作者向读者靠拢(the translator leaves the reader in peace, as much as possible, and moves the author towards him)。归化翻译法的目的在于向读者传递原作的基本精神和语义内容,不在于语言形式或个别细节的一一再现。它的优点在于其流利通顺的语言易为读者所接受,译文不会对读者造成理解上的障碍,其缺点则是译作往往仅停留在内容、情节或主要精神意旨方面,而无法进入沉淀在语言内核的文化本质深处。 有时归化翻译法的采用也是出于一种不得已,翻译活动不是在真空中进行的,它受源语文化和译语文化两种不同文化语境的制约,还要考虑到两种文化之间的

高级英语写作黄金句型和新词

1) With the rapid improvement in.../growing awareness of..., more and more.../sth.... 1 随着…的飞速发展/越来越多的关注，越来越… (e.g. With the considerable improvement in building industry, more and more structures are being erected to set the people's minds at ease.) 例：随着建筑业的大力推进，人们对建立越来越多的房屋建筑感到宽心。 2) Recently, sth./the problem of...has been brought to popular attention/ has become the focus of public concern. 2 近来，某事/某问题引起了人们的普遍关注/成了公众关注的焦点。 (e.g. Recently, the problem of unemployment has been brought to such popular attention that governments at all levels place it on the agenda as the first matter.) 例：近来失业问题引起了人们的普遍关注，各级政府已把它列为首要议程。 3) One of the universal issues we are faced with/that cause increasing concern is that... 3 我们面临的一个普遍问题是… /一个越来越引人关注的普遍问题 是… (e.g. One of the universal issues that draw (cause) growing concern is whether it is wise of man to have invented the automobile.) 例：一个越来越引人关注的普遍问题是，发明汽车是否为人 4) In the past few years, there has been a boom/sharp growth/decline in.. . 4 在过去的几年里，…经历了突飞猛进/迅猛增长/下降。 (e.g. In the past ten years, there has been a sharp decline in the number of species.) 例：在过去的几年里，物种数量骤然下降。 5) Nowadays, more/most important/dangerous for our society is... 5 如今对我们社会更（最）为重要的（危险的）事情是… (e.g. Nowadays, most dangerous for our society is the tendency to take advantage of each other in political circles.) 例：对我们社会最为危险的事情是政界倾向于互相利用。 6) According to the information given in the table/graph, we can find that... 6 根据图表资料，我们可以发现… 7) As can be seen from the table/graph/figure, there is a marked increase /decline/favorable (an unfavorable) change in... 7 根据图表（数字）显示，…明显增长（下降）/发生了有利（不利）变化。 8) As we can see from the table/graph/figure above, drastic/considerable/ great changes have taken place in...over the period of time from...(年份)to...( 年份) 8 据上面图表（数字）所示，从某年到某年某方面发生了剧烈的（相当大的；巨大的）变化。

计算机软件工程毕业设计论文

目录 目录 (1) 摘要 (1) 前言 (3) 第一章绪论 (4) 1.1研究背景 (4) 1.2设计目标 (4) 1.3本文结构 (5) 第二章系统开发环境与技术 (6) 2.1系统开发环境 (6) 2.1.1 MyEclipse插件介绍 (6) 2.1.2 Tomcat服务器介绍 (6) 2.2系统开发技术 (7) 2.2.1 JSP与Servlet技术 (7) 2.2.2 JavaScript简介 (10) 2.2.3 MVC模式 (11) 2.2.4 Struts框架 (11) 2.2.5 Spring框架 (13) 2.2.6 Hibernate框架 (15) 第三章系统需求分析与前台设计 (17) 3.1需求分析 (17) 3.1.1 系统前台简要设计概述 (17) 3.1.2 系统用例图 (18) 3.2系统设计 (18) 3.2.1 系统层次划分 (18) 3.2.2 数据库设计 (19) 3.2.3 成本管理模块时序图 (22) 第四章系统详细设计与功能实现 (27) 4.1系统项目的文件夹结构 (27) 4.2成本管理模块的具体实现 (28) 4.2.1 查询成本信息列表功能的实现 (28)

4.2.3 修改成本信息功能的实现 (36) 4.2.4 删除成本信息功能的实现 (39) 4.2.5 查看成本明细信息功能的实现 (41) 第五章总结与展望 (43) 5.1课题总结 (43) 5.2进一步开发的展望 (43) 参考文献 (44) 致谢 (45)

摘要 服饰企业生产状况联络表是针对企业的实际情况而进行设计、开发的，而成本管理模块则是为了保持产品的成本信息及时的保存、更新。利用JSP技术和SSH框架以及相应的数据库访问技术实现了基于Web的系统。该框架可以减少模块之间的耦合性，让开发人员减轻重新建立解决复杂问题方案的负担，并且可以被扩展以进行内部的定制化。通过使用JSP技术建设动态网站，充分发挥了Java语言所独有的易用性、跨平台性和安全性，从而构建了一个运行高效、安全可靠、适用性广的管理系统，实现了企业信息资源的网上管理，满足了公司业务处理的需要，使企业适应了网络经济时代发展的要求。 论文首先简要介绍了企业管理系统的一些研究与应用背景，其次介绍了该网站系统所采用的开发工具、平台以及开发环境。在此基础上，论文详尽描述了成本管理系统情况。 关键词：JSP，SSH框架，成本管理 作者：XX 指导老师：XX

本科毕业论文格式(软件工程-样例).

中文题目：物流管理系统 外文题目：LOGISTICS MANAGEMENT SYSTEM 毕业设计（论文）共××页（其中：外文文献及译文××页）图纸共 0张完成日期20××年×月答辩日期20××年×月

摘要 本物流管理系统应用于物流公司管理物流信息，主要使用了JSP、Struts、JDBC技术。控制层由Action控制流程，并调用业务层的相应方法进行不同的业务处理管理员端主要包括货物信息管理、物流信息管理、车辆信息管理、企业信息管理、客户订单管理、客户信息管理以及个人管理，管理员能对客户和货物、物流、车辆等进行增、删、查、改的操作，还能修改自己的基本信息并且在订单签订时操作员能自动提取目前登陆的用户名。客户端能查看货物车辆物流公司概况等基本信息，以及根据物流编号对物流信息进行查询，修改个人信息等操作。能更高效的提高物流公司的管理。 关键词：物流；JSP；Struts；管理

ABSTRACT The logistics management system used in logistics company management logistics information, The main use the JSP, Struts, JDBC technology. the administrator mainly includes cargo information management, logistics information management, vehicle information management, enterprise information management, management of customer orders, customer information management, and personal management, the administrator can to client and goods, logistics, vehicles and so on them, delete, check, change of operation, also can modify your basic information and signed in order when the operator can automatically extracted at present on the user name. The client can check goods vehicle logistics company profile and other basic information, and according to the logistics Numbers in logistics information query, modify the personal information and other operational. Can more efficient logistics to improve the management of the company. Keywords: logistics, JSP, Struts, management,

软件工程毕业设计说明书内容

1 引言 1.1 课题的提出 近年来随着计算机科学技术的高速发展，计算机技术也被广泛应用在我们生活的诸多领域，当然它在高校的信息化进程中也发挥着重要作用。通过先进的计算机网络技术管理高校资源，不仅提高了工作效率，而且提高了管理水平，更提高了服务质量[1]。 高校校友是一个知识体系密集、信息资源丰富、社会能力强的群体，是对自身母校有着特殊感情的群体，是潜藏在母校之外的独有的重要宝贵资源，它以桥梁和纽带的角色有效建立起学校和社会之间的联系，在学校的发展过程中发挥着重要的作用[2,3]。 1.2 课题的现状及其发展 目前世界上各种形式的校友录网站大约28万多个之多，大致有以下几类：1）以收费方式分，有收费校友录和免费校友录，其中以后者居多；2)以提供校友录服务的网站分大致有三种，有大专院校自己网站的校友录；有专门单独的校友录网站；有综合网站上的校友录；3）还有网络校友录和手机校友录之分[4]。 中国校友录发展现状：下面选取两个具有代表性的校友录来看看中国校友录网站发展的具体情况。1）中国人校友录是目前各种校友录中最具代表性、权威性的校友录。它有完善的界面服务，在校友录基本的留言、相册等功能之上，中国校友录还开通了手机校友录，同学大搜捕，星级会员等增值服务。还包括了其他信息服务内容，可以进行天气预报，股市，热点新闻的信息定制，并针对毕业班同学为他们提供全面就业信息及咨询。2）世纪同学录，现有注册用户440914人，注册班级120864个（数据截止到2004年1月10日）[5]。 1.3 本课题的主要工作 本次课题设计的是中北大学校友录管理系统，主要工作任务是实现以下系统功能：校友成员注册、登录、留言，上传并浏览照片、通讯录、系统后台管理。具体实现：校友数据的添加、修改、删除、和查询，已完成校友数据的收集及进行数据电子化；用户之间的互动，包括上传照片，相互留言，查看信息；对中北大学校友录管理系统的用户权限进行管理，以保证数据资源的合理利用。通过提供完善的校

高级英语一

北语14秋《高级英语I》导学资料一 Unit1, Unit2& Unit3 一、本阶段学习内容概述 各位同学，大家好，本课程第一阶段学习的主要内容为Unit1：Party Politics, Unit2：The New Singles, Unit3：教材课文：Doctor’s Dilemma: Treat or Let Die? 网络课件课文：Computer Violence中包括课前练习（Warm-up）、单词和词组（New words and phrases）、课文（Text）、课后练习（Exercises）及补充阅读（Supplementary readings）中的指定内容。 课前练习：大家应先了解课前练习的要求，根据已有的知识思考其中问题，或者利用网络与同学开展一些讨论，争取在阅读课文前了解文章主要论述的问题，有利于更好的了解作者的思想观点和思维过程，从而了解文章所反映的思想文化，这样既能提高阅读理解能力又能获取知识和信息。 单词和词组:名词、动词和形容词是词汇练习和记忆中的重要部分。Unit 1、2、3中所列出的新单词绝大部分都是这三类，因此，大家一定要掌握好新单词。要开发利用多种方法记单词，如联想法、音节法、构词法等。单词和词组基本上给出了英语直接释义，可以培养大家英语思维的习惯。如果在阅读释义后还有疑问，一定要查阅英语词典，寻找一些相关的解释来加深对单词的理解和记忆。许多词的释义中给出了若干同义词或近义词，能帮助大家迅速扩大词汇量，同时，课后练习的词汇（Vocabulary Study）部分又给出了一些相关的练习，大家可以在学习完单词和词组后，乘热打铁，立即做词汇练习的第一小题（选择填空）和第二小题（找出意义相近的替代词）这样可以及时巩固所学单词，大概了解这些词的用法，为正确阅读课文打下基础，也能使得练习题做起来不是那么难。 （特别说明：高级英语阶段的学习，提高词汇量和词汇运用能力是一个很大，并且很重要，同时又是比较难的问题，这是我们必须面对的问题，所以，请大家务必花时间多熟悉单词。所谓的磨刀不误砍材功，先熟悉单词和短语，才能流畅的阅读文章。更关键的是，单词和短语在考试中所占比例不少！） 课文：每单元有一篇课文（Unit1：Party Politics, Unit2：The New Singles, Unit3：教材课文：Doctor’s Dilemma: Treat or Let Die? 网络课件课文：Computer Violence）。在掌握单词和词组之后，阅读课后注释，学习课文的背景资料、作者介绍和相关内容，如人物、事件、地点等的解释，这能帮助大家准确快速的理解文章的内容。在课件资源的帮助下，认真学习课文，包括课文中常用词语和句型的用法。文章比较长，大家一定要有耐心和毅力，坚持就是胜利。在学习完整篇文章后，及时完成课文理解（Comprehension Check）的练习。其中第一小题是根据课文内容选择最佳答案，第二小题是将部分文中的句子用英语注释。只要认真学习了课件资料，相信能很快准确地完成。同时也考察大家对课文理解的程度，督促大家很好的阅读课文。 课后练习：共有四个大题。 1.课文理解（Comprehension Check）：有两个题型。在借助课件学习完课文之后，大家可以先自己做这一部分的练习，然后再看课件，对正答案的同时，再重温课文的大意。 2.词汇（Vocabulary Study）：有两个题型。在学完课文和单词后，大家可以自己先做这一部分的词汇练习，不会做得可以看课件，并牢固掌握，对于补充的单词也要掌握。这样有利于快速扩充词汇量。 3.翻译（Translation）：有的单元是汉译英，有的单元是英译汉。都是一段与课文内容相近的短文。先认真思考，仔细看文章，如果有一定的难度，可以参考课件的解说，然后再组织语言完成翻译练习。

软件工程毕业设计

天津师范大学 本科毕业论文（设计）题目：网上互动交流平台的设计与实现 学院：计算机与信息工程学院 学生姓名：龚玲玲 学号： 07509273 专业：软件工程 年级： 2007级 完成日期： 2011年5月 指导教师：夏玮

网上互动交流平台的设计与实现 摘要：在当今信息时代,计算机技术与网络技术越来越广范地应用于各个领域,改变着人们的学习、工作、生活乃至思维方式，人们越来越注重随时随地的方便快捷的交流方式，更重交流工如雨后春笋拔地而起，大大改变了人们的生活，在线交流网站应运而生，而能实现即时交互的学习平台却是凤毛麟角，总是需要麻烦的安装过程。本系统着重于用户间的交流学习，更好的迎合了广大使用者的需求。系统主要功能部分用了BS架构，只需要一台服务器，其他PC机只需要登录主机的网址便可以进入系统，进行交流，这主要归功于Jabber技术。本系统使用方便，注册简单，以用户名为关键字，不像其他交流软件，注册过程复杂，注册成功后可以根据需要自行填补信息，实现组内交流功能是一个很人性化的模块，可以根据需要实现组内成员间的“私聊”。还可以设置自己的即时状态（忙碌、在线、欢迎聊天、离线等），对于已经存在的好友也可以做不同的操作，可以聊天、编辑、添加、删除、添加分组等。 关键词：互动交流；B/S；即时

The Design and Implementation of On-line Interaction communication Platform Abstract: In this information age, computer technology and network technology is more and more widely applied in various fields, changing people's study, work and life and even a way of thinking, people pay more and more attention to the convenient anytime the way of communication, the more heavy exchange work have mushroomed ground, changed people's life greatly, on-line exchange website born, and can realize real-time interactive learning platform is rare, always need trouble installation process. This system focuses on the communication between users , better cater to the user's need. It use B/S structure on the main function part, only one server, thanks to Jabber technology, the other users just need login the host url ,then it will be able to enter the system for communication. This system is easy and comfort to use, the keyword is the unique user name , unlike other communication software,whose registration process is complicated, in this system, after successfully register ,we can fill the information as we wish, the humanized part of this system is that it achieve the goal of talking in a room. In addition, clients can set their own instant state (busy, online, welcome to chat, offline, etc.), for existing friends can also do different operation, we can chat, edit, add, delete, add group, etc. Key words: interaction；communication；B/S；in-time

软件工程毕业论文

软件工程毕业论文 Revised by BLUE on the afternoon of December 12,2020.

一、绪论 系统开发背景 随着现代社会机械化程度越来越高，人们对机械知识的渴望越来越强烈，而用户间的交流恰好满足了这种需要。用户与用户之间的互相讨论与学习会使用户快速提高自己对于机械知识的了解和认知。针对这种现状开发了本系统。 一般来说，论坛也提供邮件功能，如果需要私下的交流，也可以将想说的话直接发到某个人的电子信箱中。在论坛里，人们之间的交流打破了空间，时间的限制。在与别人进行交往时，无须考虑自身的年龄，学历，知识，社会地位，财富，外貌，健康状况，也无从知道交谈的对方的真实社会身份。这样，参与讨论的人可以处于一个平等的位置与其他人进行机械方面问题的探讨。论坛往往是由一些有志于此道的爱好者建立，对所有人都免费开放。而且，由于BBS的参与人众多，因此各方面的话题都不乏热心者。我们当然可以利用它来解决机械学习中的一些疑惑。 二、需求分析 可行性分析 可行性研究是在项目开发前期对项目的一种考察和鉴定，对拟议中的项目进行全面的、综合的调查研究，其目的是要判断项目可行与否。信息系统技术可行性研究要从系统开发的计划出发，论述系统开发力量的可行性，同时论证系统方案中所采取的各种技术手段上是否可以实现。信息系统经济可行性研究主要是对项目进行经济评价，分析系统建设投资的可能性以及评价系统运行之后给组织带来的效益。信息系统营运可行性研究要给出的方案是否可以从人力、物力、组织工作等方面保证项目按计划完成实施，还要说明项目开发后在经济、技术和环境等方面能否保证系统正常运行。 由于系统建设是一项投资大、涉及面广、工程复杂的系统工程，因此必须充分的进行可行性论证，以确保投资的准确无误，而且信息系统建设是一项整体工程，必须站在系统的角度论证它的可行性才有说服力，才有意义。可行性研究的目的是用最小的代价，在尽可能短时间内确定问题是否能够解决，它的目的不是解决问题，而是确定问题是否值得去解决，可行性从以下四个方面来考虑。 技术可行性 该课题---机械爱好者论坛，它采用了当前流行的B/S结构和Internet网络技术。而如今编写HomePage也没原来那么麻烦，网站的一些制作要求和素材在Internet随处都可以找到，制作网站的工具也是种类繁多。我们可以从中找到符合自己要求的工具。管理信息系统的开发有很多的实例，一些实例的源代码也可以提供参考。所以，从技术上来说，开发这个系统的技术难题是不多的。 三、概要设计 经过需求分析阶段的工作，系统必须“做什么”已经清楚了，现在是决定“怎样做”的时候。总体设计的基本目的就是回答“从总体上说，系统应该如何实现”这个问题，因此，总体设计又称为概要设计或初步设计。通过这个阶段的工作将划分出组成系统的物理元素------程序、文件、数据库、人工过程和文档等等，但是每个物理元素仍然处于黑盒子级，这些黑盒子

翻译中的归化与异化

“异化”与“归化”之间的关系并评述 1、什么是归化与异化 归化”与“异化”是翻译中常面临的两种选择。钱锺书相应地称这两种情形叫“汉化”与“欧化”。A．归化 所谓“归化”（domestication 或target-language-orientedness），是指在翻译过程中尽可能用本民族的方式去表现外来的作品；归化翻译法旨在尽量减少译文中的异国情调，为目的语读者提供一种自然流畅的译文。Venuti 认为，归化法源于这一著名翻译论说，“尽量不干扰读者，请作者向读者靠近” 归化翻译法通常包含以下几个步骤：（1）谨慎地选择适合于归化翻译的文本；（2）有意识地采取一种自然流畅的目的语文体；（3）把译文调整成目的语篇体裁；（4）插入解释性资料；（5）删去原文中的实观材料；（6）调协译文和原文中的观念与特征。 B．“异化”（foreignization或source-language-orientedness）则相反，认为既然是翻译，就得译出外国的味儿。异化是根据既定的语法规则按字面意思将和源语文化紧密相连的短语或句子译成目标语。例如，将“九牛二虎之力”译为“the strength of nine bulls and two tigers”。异化能够很好地保留和传递原文的文化内涵，使译文具有异国情调，有利于各国文化的交流。但对于不熟悉源语及其文化的读者来说，存在一定的理解困难。随着各国文化交流愈来愈紧密，原先对于目标语读者很陌生的词句也会变得越来越普遍，即异化的程度会逐步降低。 Rome was not built in a day. 归化:冰冻三尺,非一日之寒. 异化:罗马不是一天建成的. 冰冻三尺,非一日之寒 异化:Rome was not built in a day. 归化:the thick ice is not formed in a day. 2、归化异化与直译意译 归化和异化，一个要求“接近读者”，一个要求“接近作者”，具有较强的界定性；相比之下，直译和意译则比较偏重“形式”上的自由与不自由。有的文中把归化等同于意译，异化等同于直译，这样做其实不够科学。归化和异化其实是在忠实地传达原作“说了什么”的基础之上，对是否尽可能展示原作是“怎么说”，是否最大限度地再现原作在语言文化上的特有风味上采取的不同态度。两对术语相比，归化和异化更多地是有关文化的问题，即是否要保持原作洋味的问题。 3、不同层面上的归化与异化 1、句式 翻译中“归化”表现在把原文的句式（syntactical structure）按照中文的习惯句式译出。

高级英语写作1-10课翻译

高级英语写作1-10课翻译

The Delicate Art of the Forest 库珀的创造天分并不怎么样；但是他似乎热衷于此并沾沾自喜。确实，他做了一些令人感到愉快的事情。在小小的道具箱内，他为笔下的森林猎人和土人准备了七八种诡计或圈套，这些人以此诱骗对方。利用这些幼稚的技巧达到了预期的效果，没有什么更让他高兴得了。其中一个就是他最喜欢的，就是让一个穿着鹿皮靴的人踩着穿着鹿皮靴敌人的脚印，借以隐藏了自己行踪。这么做使库珀磨烂不知多少桶鹿皮靴。他常用的另一个道具是断树枝。他认为断树枝效果最好，因此不遗余力地使用。在他的小说中，如果哪章中没有人踩到断树枝惊着两百码外的印第安人和白人，那么这一节则非常平静/那就谢天谢地了。每次库珀笔下的人物陷入危险，每分钟绝对安静的价格是4美元/一分静一分金，这个人肯定会踩到断树枝。尽管附近有上百种东西可以踩，但这都不足以使库珀称心。他会让这个人找一根干树枝；如果找不到，就去借一根。事实上，《皮袜子故事系列丛书》应该叫做《断树枝故事集》。 很遗憾，我没有足够的篇幅，写上几十个例子，看看奈迪·班波和其他库伯专家们是怎样运用他的森林中的高招。大概我们可以试着斗胆举它两三个例子。库伯曾经航过海—当过海军军官。但是他却一本正经/煞有介事地告诉我们，一条被风刮向海岸遇险的船，被船长驶向一个有离岸暗流的地点而得救。因为暗流顶着风，把船冲了回来。看看这森林术，这行船术，或者叫别的什么术，很高明吧？库珀在炮兵部队里待过几年，他应该注意到炮弹落到地上时，要么爆炸，要么弹起来，跳起百英尺，再弹再跳，直到跳不动了滚几下。现在某个地方他让几个女性—他总是这么称呼女的—在一个迷雾重重的夜晚，迷失在平原附近一片树林边上—目的是让班波有机会向读者展示他在森林中的本事。这些迷路的人正在寻找一个城堡。他们听到一声炮响，接着一发炮弹就滚进树林，停在他们脚下。对女性，这毫无价值。但对可敬的班波就完全不同了。我想，如果班波要是不马上冲出来，跟着弹痕，穿过浓雾，跨过平原，找到要塞，我就再也不知道什么是“和平”了。是不是非常聪明？如果库伯不是对自然规律一无所知，他就是故意隐瞒事实。比方说，他的精明的印地安专家之一，名叫芝稼哥（我想，该读作芝加哥）的，跟踪一个人，在穿过树林的时候，脚印就找不到了。很明显，脚印是再也没法找到了。无论你还是我，都猜不出，怎么会

软件工程毕业设计论文

毕业设计说明书 题目：软件项目管理系统的设计和实现 系别： 专业班级： 姓名： 学号： 指导教师： 职称： 二〇一二年三月二日

摘要 在当今社会，互联网的发展，给人们的工作和生活带来了极大的便利和高效，信息化，电子化已经成为节约运营成本，提高工作效率的首选。当前大量企业的软件管理尚处于手工作业阶段，不但效率低下，还常常因为管理的不慎而出现纰漏。因此部分企业需求，设计软件项目管理系统，以帮助企业达到项目管理办公自动化、节约管理成本、提高企业工作效率的目的。 软件项目管理系统主要对项目的信息进行集中管理，方便企业建立一个完善的、强大的项目信息数据库，它是以MYSQL数据库作为开发平台。使用jsp编写程序，完成数据输入、修改、存储、调用查询等功能；并使用MYSQL数据库形成数据，进行数据存储。 软件项目管理系统是科学、全面、高效进行管理的系统，参考大量中国项目资源管理理论，根植于国内管理的实际情况，实用而科学。在操作上集输入、查询、统计等处理为一体，简便灵活，自动化功能强大。 关键字：软件管理软件项目管理系统 JSP MYSQL

Abstract In today's society, development of the Internet to the people's work and life has brought great convenience and efficiency, information technology, electronic technology has become operational cost savings, improve efficiency of choice. The current management of large enterprise employees still in the stage of manual operation, not only inefficient, but also often because of careless management flaws. So part of the business needs, design enterprise employee information management system to help companies achieve staff management office automation, saving management costs and improve work efficiency. Employee information management system is mainly focused on the information management staff to facilitate enterprises to establish a sound, strong employee information database, which is based on MYSQL database as a development platform. Programming using jsp, completion of data entry, modification, storage, call the query function; and use the MYSQL database to form data, for data storage. Employee information management system is a science, comprehensive and efficient personnel management system, reference a large number of Chinese human resource management theory, rooted in the domestic management of the actual situation, practical and scientific. The operating part one input, query, statistics and other treatment as one, easy and flexible, automated and powerful. Key words: Management System Information Management JSP MYSQL

翻译的归化与异化

万方数据

万方数据

万方数据

万方数据

翻译的归化与异化 作者：熊启煦 作者单位：西南民族大学,四川,成都,610041 刊名： 西南民族大学学报（人文社科版） 英文刊名：JOURNAL OF SOUTHWEST UNIVERSITY FOR NATIONALITIES（HUMANITIES AND SOCIAL SCIENCE） 年，卷(期)：2005,26(8) 被引用次数：14次 参考文献(3条) 1.鲁迅且介亭杂文二集·题未定草 2.刘英凯归化--翻译的歧路 3.钱钟书林纾的翻译 引证文献(15条) 1.郭锋一小议英语翻译当中的信达雅[期刊论文]-青春岁月 2011(4) 2.许丽红论汉英语言中的文化差异与翻译策略[期刊论文]-考试周刊 2010(7) 3.王笑东浅谈汉英语言中的差异与翻译方法[期刊论文]-中国校外教育（理论） 2010(6) 4.王宁中西语言中的文化差异与翻译[期刊论文]-中国科技纵横 2010(12) 5.鲍勤.陈利平英语隐喻类型及翻译策略[期刊论文]-云南农业大学学报(社会科学版) 2010(2) 6.罗琴.宋海林浅谈汉英语言中的文化差异及翻译策略[期刊论文]-内江师范学院学报 2010(z2) 7.白蓝跨文化视野下文学作品的英译策略[期刊论文]-湖南社会科学 2009(5) 8.王梦颖探析汉英语言中的文化差异与翻译策略[期刊论文]-中国校外教育（理论） 2009(8) 9.常晖英汉成语跨文化翻译策略[期刊论文]-河北理工大学学报（社会科学版） 2009(1) 10.常晖对翻译文化建构的几点思考[期刊论文]-牡丹江师范学院学报（哲学社会科学版） 2009(4) 11.常晖认知——功能视角下隐喻的汉译策略[期刊论文]-外语与外语教学 2008(11) 12.赵勇刚汉英语言中的文化差异与翻译策略[期刊论文]-时代文学 2008(6) 13.常晖.胡渝镛从文化角度看文学作品的翻译[期刊论文]-重庆工学院学报(社会科学版) 2008(7) 14.曾凤英从文化认知的视角谈英语隐喻的翻译[期刊论文]-各界 2007(6) 15.罗琴.宋海林浅谈汉英语言中的文化差异及翻译策略[期刊论文]-内江师范学院学报 2010(z2) 本文链接：https://www.wendangku.net/doc/0a18656157.html,/Periodical_xnmzxyxb-zxshkxb200508090.aspx

英语专业(高级翻译)简介

英语专业(高级翻译)简介 ·日期：2007-06-07 ·点击次数： 1446 培养目标：本专业培养具有扎实的英语语言基础，较强的语言交际能力和跨文化交际能力，掌握多方面的翻译知识和技巧，能熟练地运用英、汉语在外事、外交、经贸、文化、科技等部门从事口译、笔译的专门人才。 知识能力 要求：1．语音、语调准确，清楚自然。 2．词法、句法、章法(包括遣词造句与谋篇布局)规范、表达得体。 3．认知词汇10000~12000，且能正确而熟练地使用其中的5000~6000个单词及其最常用的搭配。 4．听、说、读、写、译技能熟练，具有较强的英语综合运用能力。毕业时达到英语专业八级水平。 听的能力：听懂真实交际场合中各种英语会话；听懂英 语国家广播电台以及电视台(如VOA, BBC, CNN)有关政治、经济、文化、教育、科技 等方面的新闻、现场报道、专题报道、综合 评述以及与此类题材相关的演讲和演讲后 的问答；听懂电视时事报道和电视短剧中的 对话。语速为每分钟150~180个单词，理解 准确率不低于60%。 说的能力：能就国内外重大问题与外宾进行流利而得体 的交流；能系统、深入、连贯地发表自己的 见解。 读的能力：能读懂一般英美报刊杂志上的社论和书评、 英语国家出版的有一定难度的历史传记和 文学作品；能分析上述题材文章的思想观 点、语篇结构、语言特点和修辞手法。能在

5分钟内速读1600词左右的文章，掌握文章 的主旨和大意，理解事实和细节。 写的能力：能写各类体裁的文章，做到内容充实，语言 通顺，用词恰当，表达得体。写作速度为30 分钟300~400个单词。能撰写长度为 5000~6000个单词的毕业论文，要求思路清 晰、内容充实、语言通顺。 译的能力：1)笔译：能运用翻译的理论和技巧，将英美报 刊上的文章、文学、科技、文化、经贸方面 的原文译成汉语，或将我国报刊、杂志上的 文章、一般文学作品、科技、文化、经贸方 面的原文译成英语，速度为每小时250~300 个单词。译文要求忠实原意，语言流畅。 2)口译：掌握连续或同步口译所需的技巧及基 本要求，能担任外经外贸、外事、外交、国 际文化、科技交流活动的口译和国际会议的 同声传译任务。 5．具有宽广的英语专业知识和相关学科知识。英语专业知识包括文学、语言学和对象国社会与文化的知识。相关学科的知识涉及教育学、教育心理学、语言习得理论、英语教学法、英语测试理论与实践、课程设置等领域。6．具有获取知识的能力、运用知识的能力、分析问题的能力、独立提出见解的能力和创新的能力。 7．熟悉中、英两种文字计算机基本技术并能实际应用。 主干课程：基础英语、高级英语、散文选读、英语泛读、英语口语、基础英语写作、高级英语写作、英汉/汉英口译、英汉笔译、英语视听说、英语报刊选读、英美文学，同声传译、连续传译、经贸法律翻译等。 修业年4年